Telco Cloud Barclays Report

Deutsche Bank
Markets Research
Industry Date
20 May 2016
Telco Cloud
Europe
Technology
Robert Sanders
Research Analyst
(+44) 20 754-58394
rob.sanders@db.com
Johannes Schaller
Research Analyst
(+49) 69 910-31731
johannes.schaller@db.com
Robert Grindle
Research Analyst
(+44) 20 754-58490
robert.grindle@db.com
Vijay Bhagavath, Ph.D

Research Analyst
(+1) 415 617-3324
vijay.bhagavath@db.com
Matthew Niknam
Research Analyst
F.I.T.T. for investors (+1) 212 250-4711

matthew.niknam@db.com
Resistance is futile Ross Seymore

Research Analyst
(+1) 415 617-3268
ross.seymore@db.com
Disruption ahead for telecom equipment vendors

Leading telco operators are embarking on major transformations to virtualize their networks. This
should enable them to emulate the agility of over-the-top (OTT) providers and save on cost. While
progress to date has been slow, we believe momentum should build over the next 3 years. This
could raise questions over the traditional model of the equipment vendors. We see Nokia's
strengths coming to the fore and providing resilience, and justifying a premium to Ericsson. We
also see positive implications for ARM's processor architecture given gains in networking from
Power and MIPS (owned by Imagination).
________________________________________________________________________________________________________________
Deutsche Bank AG/London
Deutsche Bank does and seeks to do business with companies covered in its research reports. Thus, investors should
be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should
consider this report as only a single factor in making their investment decision. DISCLOSURES AND ANALYST
CERTIFICATIONS ARE LOCATED IN APPENDIX 1. MCI (P) 057/04/2016.
Deutsche Bank
Markets Research
Europe Industry Date

20 May 2016
Technology
Telco Cloud
FITT Research
Robert Sanders
Resistance is futile Research Analyst

(+44) 20 754-58394
rob.sanders@db.com
Disruption ahead for telecom equipment vendors Johannes Schaller

Leading telco operators are embarking on major transformations to virtualize Research Analyst
their networks. This should enable them to emulate the agility of over-the-top (+49) 69 910-31731
(OTT) providers and save on cost. While progress to date has been slow, we johannes.schaller@db.com
believe momentum should build over the next 3 years. This could raise
questions over the traditional model of the equipment vendors. We see Nokia's Robert Grindle
strengths coming to the fore and providing resilience, and justifying a premium
Research Analyst
to Ericsson. We also see positive implications for ARM's processor architecture
(+44) 20 754-58490
given gains in networking from Power and MIPS (owned by Imagination).
robert.grindle@db.com
Network virtualization is entering the deployment phase
Our industry conversations suggest that leading telco operators are continuing Vijay Bhagavath, Ph.D
to push ahead with network virtualization, attracted by cost savings and Research Analyst
increased agility. Significant obstacles remain, such as interoperability, lack of (+1) 415 617-3324
skills and legacy hardware. However, “innovator” operators (which comprise vijay.bhagavath@db.com
50% of global capex) are showing determination with initial deployments in
contained domains (eg vEPC, vCPE). Telcos are increasingly attracted by new
Matthew Niknam
revenue opportunities focused on new industry verticals, enabled by NB-IoT
and 5G, which have virtualization at their foundation. Research Analyst
(+1) 212 250-4711
Software is set to disaggregate from hardware matthew.niknam@db.com
Virtualization disaggregates software from dedicated hardware, posing a
challenge to equipment vendors. To date, their business model has been to up-
Ross Seymore
sell higher-margin, proprietary software-led upgrades into an installed base of
lower-margin hardware. Virtualization compels vendors to deliver software- Research Analyst
only offerings on standard hardware. This opens up a concentrated market to (+1) 415 617-3268
new entrants. Perhaps it is no surprise that vendors have been guilty of ross.seymore@db.com
spreading FUD (fear, uncertainty and doubt). However, we believe that
“resistance is futile” for vendors, and could hurt competitiveness. Top picks
We continue to see Nokia (Buy) meriting a 15% premium over Ericsson (Hold) Nokia (NOKIA.HE),EUR4.61 Buy
In the face of this challenge, we believe that Nokia should remain on the front Source: Deutsche Bank
foot, embracing open standards and leveraging its superior R&D scale (35,000
Companies Featured
R&D heads post restructuring). The acquisition of Alcatel-Lucent has given
Nokia unrivaled breadth (access, routing, optics), and key cloud assets Nokia (NOKIA.HE),EUR4.61 Buy
(eg Cloudband, Nuage). By contrast, Ericsson’s services-led approach seems Ericsson (ERICb.ST),SEK62.45 Hold
more vulnerable from the transition to cloud, while its partnership with Cisco is ARM Holdings (ARM.L),GBP925.00 Hold
muddled. We continue to believe Nokia merits a 15% premium to Ericsson, Imagination (IMG.L),GBP165.50 Sell
versus a 7% discount today, (EV/EBITDA 2018E: Nokia 4.8x, Ericsson: 5.1x). Spirent (SPT.L),GBP81.25 Hold
ADVA (ADAG.DE),EUR8.38 Hold
We up our target on ARM (Hold) to £10.50. We downgrade Imagination to Sell Source: Deutsche Bank
With the smartphone and PC markets now mature, next-generation
networking (along with auto) remains one of the few growth drivers left in the
semiconductor industry. We believe that ARM should continue to take share
against the MIPS (owned by Imagination) and Power architectures, with
ARM’s unit share in networking set to rise from 15% today to 51% by 2020.
We upgrade our target on ARM to £10.50 (from £10), but keep our Hold rating
given concerns on mobile royalties, which are still much more significant than
networking. We downgrade our rating on Imagination Tech from Hold to Sell
with a 120p target (27% downside), given declining MIPS royalties and high
exposure in graphics IP to weak demand trends at Apple.
________________________________________________________________________________________________________________
Deutsche Bank AG/London
Deutsche Bank does and seeks to do business with companies covered in its research reports. Thus, investors should
be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should
consider this report as only a single factor in making their investment decision. DISCLOSURES AND ANALYST
CERTIFICATIONS ARE LOCATED IN APPENDIX 1. MCI (P) 057/04/2016.
20 May 2016
Technology
Telco Cloud
Table Of Contents
Executive summary ............................................................. 3

The rising importance of the telco (network) cloud ............................................ 3
Disruption ahead for equipment vendors ........................................................... 3
Implications for coverage in network equipment ............................................... 5
Telcos are set to benefit from network supplier disruption ................................ 6
The ARM processor architecture should benefit ................................................ 7
Telcos: the story so far ........................................................ 8

Key points ........................................................................................................... 8
Telcos: traffic growth is pressurizing profitability ............................................... 8
The cloud has highlighted telcos’ lack of agility ............................................... 10
5G and IoT to create new enterprise opportunities for telcos ........................... 13
Telco cloud: new paradigm ............................................... 14

Key points ......................................................................................................... 14
Telco network cloud – an emerging paradigm ................................................. 14
Two key concepts - NFV and SDN.................................................................... 15
Key benefits and barriers to adoption ............................................................... 17
Operator engagements to date ......................................................................... 20
Focus on mobile networks ................................................ 22

Key points ......................................................................................................... 22
The LTE network architecture today ................................................................. 22
Virtualization impact on mobile networks ........................................................ 23
Operator and vendor views diverge on the pace of change ............................. 27
Implications for vendors .................................................... 30

Key points ......................................................................................................... 30
1) Reduced vendor lock-in – vendors and operators at odds ............................ 30
2) Software unlikely to mitigate lost profit entirely ........................................... 31
3) Increased competition from IT vendors and startups ................................... 31
4) Services – a mixed picture ............................................................................ 32
5) Organizational disruption is ongoing ............................................................ 34
6) Enterprise channel is growing in importance ............................................... 36
7) Financials: Impact initially small, but becomes larger .................................. 37
Implications for coverage .................................................. 42

Key points ......................................................................................................... 42
Some initial conclusions ................................................................................... 44
Implications for the European network equipment universe ............................ 44
Implications for the US networking equipment universe .................................. 46
Implications for Telcos ...................................................... 48

Positive impact on telecom operators from network cloud .............................. 48
Implications for ARM ........................................................ 51

Key points ......................................................................................................... 51
Networking in the processor architecture landscape ....................................... 51
ARM is likely to take a big share in networking ................................................ 53
Competitive environment - a “battle royale” is looming ................................... 56
We increase our target price on ARM to £10.50 .............................................. 61
We downgrade Imagination Technologies to Sell ............................................ 62
Reference architectures..................................................... 64
Glossary ............................................................................. 66
Acronyms ......................................................................................................... 67
Page 2 Deutsche Bank AG/London

20 May 2016
Technology
Telco Cloud
Executive summary
We believe that this is the first comprehensive report for the investment
community that covers an important upcoming disruption, namely the “telco
network cloud”, and should be of high interest to long-term investors in both the
technology and telecom sectors. While we are at the beginning of a lengthy
transition, the impact of this trend will be far-reaching, and we expect the key
themes to materially drive investment cases over the medium-term. There are
numerous sectors in TMT that could be impacted including network equipment
vendors, telecoms, and semiconductor companies. We touch on all of these in
the following FITT report. Please do not hesitate to reach out with questions.
The rising importance of the telco (network) cloud
Telcos are continuing to see high levels of traffic growth driven by video (eg
Youtube, Netflix), but monetization is continuing to prove a challenge, and
profitability continues to decline. The advent of cloud concepts inside telecom
networks like NFV and SDN, what we call the “telco network cloud” or “telco
cloud” for short, offers an opportunity to address the challenges being posed
by OTT (over the top) vendors like Amazon, which can launch a service in a
matter of seconds (“failing fast”) versus several months for a telco to launch its
services. In addition, we believe that full adoption of NFV and SDN can drive
opex and capex savings of up to 35% and 20% respectively.
While this may sound like “future music”, large tier-1 operators are starting to Figure 1: Customer concentration
embrace virtualization, as part of a 10-year transition. John Donovan, from first
mover AT&T, recently talked about their target to virtualize 75% of its network Top 10 customers as a % of sales
functions by 2020 as a matter of “survival”. That kind of commentary does not
46% 50%
relate to worries about intra-telco competition, but is a response to the direct
threat posed by the OTTs. This becomes even more the case in a “multi-play”
world where fixed-line telcos, wireless telcos and pay TV companies are
increasingly the same company. Further, new revenue opportunities from
internet-of-things (IoT) and 5G will likely have virtualization at their foundation. Ericsson Nokia
These tier-1 operators matter because they comprise 50% of global capex, and
a similar percentage of revenue for our coverage (see Figure 1). Source: Company filings, Deutsche Bank estimates
Disruption ahead for equipment vendors
Network functions virtualization (NFV) de-couples hardware and software Figure 2: Ericsson: business mix
business and has a meaningful impact for equipment vendors, which up until
today have sold dedicated hardware as the primary entry point. The business
model for the likes of Nokia and Ericsson has been to up-sell proprietary,
software-led upgrades into an installed base of lower-margin hardware, not
Hardware
dissimilar to the classic “razor blade” model. If the hardware becomes a Services
34%
43%
commodity (an off-the-shelf server), the stickiness or “vendor lock in” that
vendors enjoy becomes less meaningful, while new entrants can enter the fray.
Software
23%
Vendors have responded by participating in proof of concepts (PoCs) around
NFV and SDN, while demonstrating reluctance to fully participate in the
development of open standards, the basis of the new model. These challenges Source: Company filings
of interoperability have hindered progress with the new technology. In
addition, vendors have been accused of spreading fear, uncertainty and doubt
Deutsche Bank AG/London Page 3

20 May 2016
Technology
Telco Cloud
(FUD) around benefits. Leading telcos like AT&T, Telefonica, Vodafone and
Verizon are being forced to take matters into their own hands, driving adoption
of open standards, and exploring use of startups. Vodafone just this week
revealed a target of virtualizing 50% of network functions by 2020. So for
vendors, we believe that “resistance is futile”, given that some are attempting
to slow the pace of change to their traditional business model.
So while significant challenges remain for telcos to fully adopt NFV and SDN
concepts in many parts of the network, we believe that we are now entering
the “deployment phase” and that momentum should build over the next 3
years. This should increase the profile of telco cloud on the investment cases
of many companies. Given our coverage in Europe, we focus this report on the
impact on Nokia and Ericsson amongst the equipment vendors, but there are
clearly implications for a multitude of companies across hardware, software
and IT services. We highlight them in Figure 3.
Figure 3: Tech companies’ exposure to telcos (hardware, software, services)
Telco-exposure on total sales
Ericsson 90%
Nokia 87%
Ciena 87%
Amdocs 80%
ZTE 66%
Juniper 61%
Huawei 58%
Cisco 20%
NEC 17%
Accenture 10%
Capgemini 7%
Fujitsu 5%
Oracle 5%
IBM 3%
HP Enterprise 3%
Samsung 2%
Source: Deutsche Bank
C-RAN could be more impactful in mobile networks near-term

When we look more closely at mobile networks given their high relevance for
Nokia and Ericsson, current virtualization efforts are focused on the mobile
packet core, which on the face of it is a small fraction of equipment spend at
just 4%. However, cloud-like, but not “full” virtualization concepts like Cloud
Radio Access Networks (C-RAN) should be impactful between now and 2020.
C-RAN allows for pooling of baseband resources and offers up significant
capex savings, and will take up more of the core base station revenue
opportunity over time, reaching 33% of mobile infrastructure market
(ex services) by 2020. Full virtualization of the RAN (radio access network)
looks further out, but remains a potential future threat from 2022 onwards. So
while weak wireless capex trends (-7% in 2016 and -5% in 2017) led by
tapering LTE rollouts are more meaningful today for the investment cases of
wireless equipment vendors, we see telco cloud rising in importance over time.

20 May 2016
Technology
Telco Cloud
We show below in Figure 4 above our overall view on the top-line impact on
wireless equipment vendors. We see the net negative impact from NFV and
SDN to become meaningful from 2018 onwards, and rising to a -20% negative
impact by 2025, relative to the existing run-rate of wireless equipment spend
(which is declining by 2% CAGR to 2019), and currently driven by the tapering
off of LTE rollouts. Overall, we see increased software sales from NFV/SDN,
but we don’t believe this can fully compensate for lost hardware sales at the
profit level, given price aggression of new entrants, and potential adoption of a
subscription model. We believe that services should see a positive opportunity
from transformation opportunities, but a negative impact from reduced
managed services and network rollout sales.
Figure 4: Likely impact of telco cloud on the wireless infrastructure market
Adoption curve
2014-15 2016-17 2018-19 2020-23 2024-25
Proof of Early Early Broad Laggards

concepts adopters majority adoption
Key domains Mobile Core (vEPC, vIMS etc), C-RAN vRAN
20%
 Uncertainty around roadmap  vRAN
Net revenue +  Virtualized mobile core  Subscription-model
impact on
 C-RAN savings led by startups
wireless 0%
infrastructure  Managed
market including -  Consulting and SI uplift services decline
services ($100bn)
 White-label Intel hardware sales
-20%
Implications for coverage in network equipment
European coverage: we continue to prefer Nokia over Ericsson

We prefer Nokia (Buy, €6.2 target) over Ericsson (Hold, SKr69 target), given
Nokia’s unrivaled breadth (radio, core, routing, optics, fixed) following the
acquisition of Alcatel-Lucent. We also believe that Nokia’s positions in C-RAN,
NFV and SDN are market-leading. In addition, our view is that Nokia’s lowly
valuation already discounts the sluggish environment for LTE rollouts, while
insufficiently discounting the software growth opportunity to come, led by
software-led upgrades to LTE Advanced and LTE Advanced Pro. On valuation,
we believe that Nokia’s merits a 15% premium to Ericsson, versus a 7%
discount today, (EV/EBITDA 2018E: Nokia 4.8x, Ericsson: 5.1x).

20 May 2016
Technology
Telco Cloud
By contrast, Ericsson (Hold, SKr69 target), appears to be pushing a services-led

approach (“we’ll do everything for you”), which appears defensive and more
vulnerable from the transition to the cloud which empowers operators to pick
and choose vendors, while changing the required skillset. We have also picked
up signs that Ericsson is more resistant to adopting open standards, which is a
concern. Finally, we also believe that Ericsson’s partnership with Cisco is
muddled, given their product overlap, while the partnership’s sustainability is
uncertain, given potential disagreements and management change at Ericsson.
US networking equipment: we see a positive read for Cisco and Juniper

In our view, the switching and routing vendors - Cisco and Juniper (both Buy By Vijay Bhagavath & team
rated), have a multi-year thematic opportunity in telco/carrier transformations,
driven by the migration to a software-based and highly automated services
delivery infrastructure. A core software-based network services delivery and
automation framework is at the center of any NFV cloud initiative.
Cisco and Juniper, in our view, are in the early innings of getting design wins
for their cloud automation and service orchestration frameworks - e.g. Juniper
Contrail, Cisco APIC (Application Policy Infrastructure Controller), as the core
SDN and NFV platform - for enabling automated and dynamic delivery of
network services such as business VPNs, Ethernet and optical connectivity,
managed security services etc.
Our industry conversations with the telco/carrier IT channel suggest that Cisco
is in the early innings of selling turnkey telco cloud solutions for web-based
services ordering and automated fulfillment for network service offerings such
as layer 3 business VPNs, Ethernet and optical connectivity, managed security
services, etc. These turnkey engagements are typically multi-year opportunities
for Cisco, involving sales of edge and core routers, metro optical, software
automation and service cataloging solutions, and software- based mobile
packet core, IP video delivery etc.
For Juniper, our view is very similar to Cisco. We believe Juniper is leading the
customer dialogue with their Contrail NFV cloud software automation solution
- so as to pull forward multi-year sales for their edge routing, switching and
network firewall platforms, and related network integration opportunities. The
publicly announced Contrail design win at AT&T Domain 2.0 for NFV is a case
in point.
Bigger picture, we see Cisco and Juniper, in particular, increasingly leading

with their telco cloud automation frameworks to create a "sticky" multi-year
order pipeline for their networking platforms at the major telcos and cable
companies over the next few years
Vijay has written on the transition at networking companies to software in a

recent FITT report, published on April 10tth.
Telcos are set to benefit from network supplier disruption
The DB telecom services research team views the ‘cloudification’ of network By Robert Grindle & team
resources and the associated disaggregation of hardware and software supply
as a welcome continuation of an ongoing process of efficiency gains for the
industry, akin to switch digitalization, fiber optics, or in purely economic
terms, the ascendance of Chinese vendors. We believe that efficiencies in

20 May 2016
Technology
Telco Cloud
signal coding and greater spectrum availability and Moore’s Law have
materially reduced the unit costs of deploying fixed and mobile data capacity.
We estimate that mobile operators like Vodafone are already seeing unit
capacity costs falling in excess of 30% annually. This is just as well given high
levels of consumer demand and the need to offset legacy voice revenues. All
telcos should benefit from increased efficiency gains, but larger operators, and
those with cross-border footprints, may prove the biggest beneficiaries. In
addition, new revenue opportunities from IoT, which have virtualization at their
foundation could also become signficiant on a long-term view.
The ARM processor architecture should benefit
We see the disruption to the equipment supply chain as positive for the ARM
processor architecture given that the large majority of key semiconductor
companies have licensed ARM’s v8 core for networking to enable next-
generation designs supporting SDN, NFV, C-RAN and 5G. We expect almost all
networking semiconductor companies except Intel to ramp up ARM-based
products between now and 2020, increasing ARM’s unit share from 15% in
2015 to 51% in 2020 (41% by value), at the expense of Power and MIPS
(owned by Imagination). We raise our target on ARM to £10.50 (from £10), but
keep a Hold rating, while we downgrade Imagination to Sell (120p target, 26%
downside), given loss of share in MIPS and declining GPU royalties from Apple.
Figure 5: Market share shifts in networking (by value)
45%
41%
34%
ARM
Intel
25%
19% MIPS
8% 12% Power
8% Others
2014 2016E 2018E 2020E
Source: Deutsche Bank; Note Power is an architecture that was originally born out of IBM technology
Intel is also set to see traction in networking

We also see Intel as well positioned to gain significant share in networking,
driven by the advent of NFV and SDN. Furthermore, its acquisition of Altera
has the potential to fast-forward Intel’s position in the data plane.

20 May 2016
Technology
Telco Cloud
Telcos: the story so far

Key points
 Telcos are being over-whelmed by video traffic, growing at 59% CAGR
 Annual cost reductions per bit of 20-30% are struggling to keep up
 4G investments have also not averted a decline in ARPUs
 Telco’s EBITDA margins have thus seen a steady decline
 Implementing cloud concepts in the network over time could create
help to reduce opex and capex, and create new revenue opportunities,
notably driven by 5G and IoT (internet-of-things)
Telcos: traffic growth is pressurizing profitability
Telcos are seeing fast-rising growth in IP traffic, driven primarily by mobile.

Mobile and WiFi offload from mobile devices will account for almost 30% of
total IP traffic by 2019. By contrast, traffic over fixed networks is set to
continue to grow, although at a much slower pace. The main driver for the
explosion of mobile traffic (59% CAGR to 2019) is video, which already
represents 55% of mobile data, and will likely represent 75% in 2020.
Figure 6: IP traffic by access technology, worldwide, exabytes per month
180 22% CAGR 2014-2019

Fixed/Wired (11% CAGR)
160 Fixed/Wi-Fi from Wi-Fi-Only Devices (22% CAGR) 29% from
13%
Fixed/Wi-Fi from Mobile Devices (67% CAGR) Mobile
140 Devices
Mobile Data (59% CAGR) 16%
120
53% is
100 Wi-Fi
Exabytes
per month 80 37%
60 4%
3%
40 38%
34%
20 55%
0
2014 2015 2016 2017 2018 2019
Source: Cisco VNI Mobile, 2016
The increase in data growth necessitates huge capital investment by telcos,

but to date this has not been accompanied with rises in ARPUs (average
revenue per user). In fact, the global trend for ARPU has been for gentle
declines of around 3% pre annum, so significant reductions in cost per bit have
been necessary just to maintain the telco profit pool. So far, the likes of
Facebook, Skype, and Whatsapp, have already led to service providers losing
$14bn in lost voice and messaging revenues in 2014, according to Juniper
Research, but monetizing video growth is the main challenge looking forward.

20 May 2016
Technology
Telco Cloud
Mobile operators - questionable payoff on 4G investments

Given our coverage of Ericsson and Nokia, we mainly focus on mobile
operators in this report. Mobile operators have spent around $700bn globally
on network infrastructure in the decade to 2015. In the early part of the
decade, that money went on 3G, while from 2009 onwards, 4G spending
began, which peaked in 2015 as large rollouts in key geographies (US, Japan,
China) have since slowed. At the same time, the launch of the iPhone has led
to radically new data usage patterns and large capacity requirements.
However, this has not led to revenue growth – on the contrary, as can be seen
in Figures 7 and 8, revenue growth has remained sluggish for telco operators.
In fact, the median revenue growth is just 1-3%, despite the advent of 4G. Not
only have telcos struggled to push through ARPU increases, but cost per bit
reductions have struggled to keep pace with traffic growth, leading to EBITDA
margins declining over the 10 years from 2005 to 2015, with declines larger for
the largest 10 operators.
Figure 7: Telco growth rates (%), 2005-15 Figure 8: Telco EBITDA margins (%), 2005-15
Median growth rate (top 10 operators) Median EBITDA margin (top 10 operators)
Median growth rate (sample of 40 operators) Median EBITDA margin (sample of 40 operators)
20% 45%
15%
40%
10%
35%
5%
0% 30%
2005 2007 2009 2011 2013 2015 2005 2007 2009 2011 2013 2015
Source: Deutsche Bank Source: Deutsche Bank
According to the Ericsson Mobility Report (Nov’15), Youtube accounts for

between 50-70% of total video traffic of mobile networks, regardless of
terminal type. Netflix represents 10-20% of mobile video traffic in countries
where they have launched. Sadly for telcos, net neutrality principles limit their
ability to monetize excessive data usage by heavy users of Youtube and
Netflix, although that is not stopping telcos from trying (see Comcast’s recent
attempt to deal with the 8% of customers on their network that consume in
excess of 300GB/month, and the protests that have led them to raise this
treshold to 1TB/month). This was seen as a way to mitigate the impact of
“cord cutting” on Comcast’s pay TV services, given that in-house services
conveniently don’t apply to the cap.
Next-generation technologies are essential to reduce cost

Nokia’s Bell Labs Consulting division published a report in April 2016 on the
traffic burdens that mobile operators are facing. Their work suggests that
adoption of next-generation technologies can drive an annual cost per GB
decline of 20-30% per year, up to a maximum of 30% per year, assuming full
adoption. This can be seen in Figure 9.

20 May 2016
Technology
Telco Cloud
Nokia then go on to calculate the level of data growth that operators can
support profitably by region – see Figure 10. This varies from 32% to 51%
compound growth. The current issue is that this is below the 55-79%
compound growth currently seen at operators. As a result, there is a “demand
gap” relative to the capacity that can profitably be supported. Telcos then will
have to use a combination of tools: data caps, increased prices and cost
reduction activities to maintain current levels of profitability. The recent trend
of declining profitability suggests this has been tough to achieve. However,
Nokia’s work suggests that those countries like North America which have the
greatest enthusiasm for next-generation technologies (5G, NFV/SDN) will be
able to support the highest consumption per device, given the increased cost
reduction this should support.
Figure 9: Cost trends (North America), per GB Figure 10: Profitable cellular consumption per device
(GB)
$14 16
North
America:
$11.44
14 51% CAGR
$12
CAGR: 2014-20E 12
$10 Developed Asia
Maximum of a 30%
10 Pacific:
annual reduction in 34% CAGR
$8
cost per GB
Small cells 8 China
35% CAGR
$6
6 Europe
32% CAGR
$4
4 Latin America
SDN/NFV 41% CAGR
$2 $1.35
2 Emerging
5G, LTE-U Asia/MEA
$0 0 38% CAGR
2013 2014 2015 2016 2017 2018 2019 2020 2014 2015 2016 2017 2018 2019 2020
Source: Deutsche Bank, Nokia Bell Labs Mobility Traffic Report Source: Deutsche Bank, Nokia Bell Labs Mobility Traffic Report
The cloud has highlighted telcos’ lack of agility
Figures 11 and 12 illustrate the important point about telcos’ lack of agility.
Telcos have been working on a timeline of 2-6 years when it comes to
launching new services (which could range widely from testing out
promotions, new video services, the launch of a MVNO or a tailored customer
offering for a one-off event). This is because to date the telco deployment cycle
has been conditioned by hardware.
This hardware has been complex and expensive, while by contrast, cloud
service providers have been quick to launch services. Over time, this has led to
telcos looking slow and cumbersome compared with the much nimbler cloud
providers, which are able to “fail quickly” and innovate. The good news is that
telcos are now embarking on a major and challenging transformation to adopt
the cloud paradigm, which could ultimately lead to innovative services, and
potentially telcos’ products and services actually being prioritized over OTT
applications, were that to pass the scrutiny of the regulators.

20 May 2016
Technology
Telco Cloud
Figure 11: Telco cycle Figure 12: Cloud cycle
Source: Telefonica Source: Telefonica
Cloud has already disrupted the IT environment

Cloud computing continues to be the key disruptive trend in the IT industry. Virtualization in an IT context
Cloud computing provides shared processing resources and data to computers refers to creating a virtual
and other devices on demand. The main enabling technology of cloud is version of a resource (server,
virtualization. Virtualization software separates a physical computing device
OS, data storage device,
into one or more “virtual” devices, each of which can be easily used and
network service), which is not
managed to perform computing tasks. Idle computing resources can be
the physical device itself that
allocated and used more efficiently. Virtualization provides agility to flexibly
speed up IT operations, while reducing costs by increasing infrastructure both behaves and appears to
utilization. the user as the actual device
The rise of “hyperscale” cloud service providers reflects the significant benefits
to scale that can be realized from building large data centers with hardware
tailored to the specific need of that provider. Revenue for leaders in public
cloud like Amazon Web Services (AWS) has exploded accordingly (see
Figure 13). Amazon gives its customers access to an entire shared pool of
compute resources, first with core server and storage capacity, and then up
the stack to database and even certain application software. Many enterprises
like GE, News Corp, Time, Coca-Cola and Netflix are moving large portions of
their IT workloads to AWS. Deutsche Bank has written research on this subject
(see the FITT report written on AWS by US software analyst, Karl Keirstead,
from November 2015).

20 May 2016
Technology
Telco Cloud
Figure 13: Amazon Web Services, revenue ($bn) and yoy growth (%)
70%
49% 49%
30% Amazon Web
Services
(AWS)
15.3 revenue
11.7
7.9 yoy growth
4.6 (%)
3.1
2013 2014 2015 2016E 2017E
Source: Company financials, Deutsche Bank estimates
In contrast to public cloud, private cloud is cloud infrastructure operated solely

for a single organization, whether managed internally or by a third party
outsourcing provider like IBM, HP, CSC or Infosys. Cost savings and efficiency
gains are again realized by an enterprise centralizing and standardizing their IT
infrastructure in fewer, but more modern data centers. Private cloud offers
reduced cost savings, but allows data integrity not be compromised.
Traditional IT services companies have been net losers

While the likes of AWS have grown to fantastic scale, the transition to the
private cloud from legacy IT has not benefited traditional IT services companies
like HP, IBM and CSC, while their attempts to mimic Amazon and Microsoft
Azure in public cloud have failed (eg HP). In fact, most asset-heavy
infrastructure IT service companies have seen a net negative impact from the
cloud given reduced implementation spend, less customization, and reduced
outsourcing deals to manage legacy IT. This is despite the potential
opportunities that arise from system integrations expertise and potentially
custom application work. Our US IT Services analyst, Bryan Keane, has written
a series of FITT reports on this topic in May 2015, and another in May 2016.
Figure 14: IT infrastructure outsourcing firms, TTM revenue ($bn), yoy growth
CSC
-3% -3%
-5% -6%
-7% -9% -11% -11%
Hewlett
Packard
Enterprise
55 55 54 53 52 50 48 47 IBM
aggregate
yoy growth
(%)
Q1-14 Q2-14 Q3-14 Q4-14 Q1-15 Q2-15 Q3-15 Q4-15
Source: Company financials

20 May 2016
Technology
Telco Cloud
Most “cloud” ventures at telcos to date have been in “public cloud”

Many telcos are not complete strangers to cloud computing, but more in the
sense of providing external services rather than transforming their networks. In
fact, many fixed-line telcos saw the rise of public cloud as an opportunity to
provide their own cloud solutions to large and small enterprises, both in
infrastructure and in applications. Most notably, Verizon acquires Terremark
for US$1.4bn in 2011. Terremark had 200 data centers with offices in 75
countries. Other telcos that have tried to play here include NTT, AT&T, Orange,
BT, and Deutsche Telekom, but success has been limited to date. In a swift
about turn, telcos like Verizon, Telefonica, Centurylink and Windstream have
been putting “for sale” signs up on many of their data centers, while AT&T
recently transferred its managed hosting business to IBM. In our view, this was
because they lacked the scale to compete with AWS-like price points. Perhaps,
the telcos with the best chance of success in public cloud are those that can
leverage concerns around data integrity/privacy in local markets (eg Deutsche
Telekom in Germany).
The telcos have since re-focused their cloud efforts away from classic IaaS
(infrastructure-as-a-service) towards more “stickier”, value-added services for
small-and-medium-sized enterprises (SMEs), for example security and mobility
services. Telcos are also now leveraging their large customer bases to provide
network access to the large third-party cloud providers like AWS and Microsoft
Azure. In addition, telcos are now looking to provide WAN optimization
functions that were previously provided by appliances as cloud services.
5G and IoT to create new enterprise opportunities for telcos
Although US operators are already “talking up” 5G, the 5G standard setting
process won’t be complete until 2020 (release 16). However, it is already clear
that 5G and LTE-Advanced Pro will be driven as much by new use cases, as it
will be by faster mobile broadband for typical consumer video use cases. Many
of these new use cases will be driven by specific vertical requirements around
latency, connection density and throughput which will be inherently
contradictory. This will create the need for “network slicing”, tailored networks
for specific applications (eg automotive, public safety, healthcare) that leverage
cloud concepts like NFV and SDN which we will discuss in the next section.
Figure 15: Challenges and gaps to reach 5G

Latency Throughput Mobility Connection Density Network architecture
5G 1ms 10GBps 500 km/h 1,000k Slicing
E2E latency Per connection High-speed railway Connections per km2 Ability required
Use case Self-driving cars Mobile broadband Mobile broadband M2M communications Vertical-focused
Already supported with
LTE-Advanced Pro (rel13)
Gap to LTE 30-50x 100x 1.5x 100x NFV/SDN
LTE 30-50ms 100Mbps 350 km/h 10k Inflexible
Source: Deutsche Bank, ETSI
Furthermore, we see narrowband-IoT (NB-IoT) as the most likely candidate to

be the connectivity standard of choice (ahead of Sigfox, LoRa et al). This
technology should be available in 2018, as part of release 13 of the standards
(LTE-Advanced Pro). So telcos will not have to wait to 2020 before being in a
position to leverage IoT as a meaningful revenue source. AT&T, Verizon,
Telefonica, Deutsche Telekom, KPN, Orange, China Mobile are all early movers
here. Cloud concepts will play an increasingly important role here in our view.

20 May 2016
Technology
Telco Cloud
Telco cloud: new paradigm

Key points
 Implementing cloud concepts in the network architecture aims to
emulate the agility of cloud providers like Amazon. At the same, time
telcos are attracted by the promise of reduced opex and capex
 Network functions virtualization (NFV) enables network resources to
be shared dynamically on commodity hardware (x86 servers)
 Software-defined networking (SDN) is complementary, and allows
control of the network to be de-coupled from the physical hardware
 Significant barriers remain to broad-based adoption: including skills
shortages, interoperability, and integration with legacy equipment
 Early mover operators include AT&T, Telefonica, Verizon and DT.
Popular early use cases for virtualization include vEPC and vCPE.
Telco network cloud – an emerging paradigm
True “telco network cloud” promises to radically change the telecom industry
landscape by introducing cloud computing concepts into the network
architecture itself. This involves the creation of a common virtualized
infrastructure to deploy and operate different network applications, provided
by multiple network vendors. The common infrastructure can be built with
COTS (commercial off the shelf) data center equipment, enabling capex and
opex savings, whilst the time to launch services can be radically shortened.
Figure 16: Amazon: lessons in efficiency and agility Figure 17: Value of the telco cloud
13%
30%
16%
16%
25%
Shorter innovation cycles for deploying new services

Cost reduction
Cope with fluctuating customer or service demand
New revenues by cloud services
Better experience for customers
Source: Nokia Source: Nokia operator survey

20 May 2016
Technology
Telco Cloud
Two key concepts - NFV and SDN
Two key concepts under-pin the emergence of telco cloud:: NFV and SDN.
NFV (network functions virtualization), and SDN (software-defined networks).
NFV was created by a consortium of telcos in 2012 that were frustrated that Network virtualization enables
existing hardware had limited their ability to speed up deployment of new network resources to be
services. NFV is then an initiative to virtualize the network services that are shared dynamically by
now being carried out by proprietary, dedicated hardware. NFV disaggregates
dividing the network’s
software and hardware with network functions being run in software via
available bandwidth into
virtualization techniques and then runs them on commodity hardware (today
separate, individual channels
this is standardized Intel-based, x86 servers). This allows virtual appliances to
be launched on demand without the delays that come from installing new where each can operate
equipment. NFV then offers the promise of allowing operators to increase independently
resource utilization, reduce hardware purchasing costs, reduce opex and most
importantly launch services more rapidly. In Figure 19, we illustrate the impact
on the classical network appliance approach.
Figure 18: Classical network appliance approach Figure 19: Network functions virtualization (NFV)
approach
 Fragmented non-commodity hardware

 Physical install per appliance per site
 Hardware development large barrier to entry
for new vendors constraining innovation &
competition
Source: ETSI Source: ETSI
SDN by contrast was born in the data center, and is a networking technology SDN is a networking model
that decouples the control plane from the underlying data (aka where control of the network
user/forwarding/bearer) plane and consolidates the control functions in a is decoupled from the
centralized controller. SDN is thus an approach to networking that allows
physical hardware
administrators to manage a network through the abstraction of “higher-level”
functionality. In a classic SDN scenario, rules for packet handling are sent to
OpenFlow is a standard
the switch from a controller. Controllers and switches normally communicate
via the “southbound” interface, normally the OpenFlow application protocol between centralized
programming interface (API) protocol. Northbound interfaces interact with control point and network
applications via APIs. SDN architectures are characterized by being open, switches.
directly programmable, configurable, agile, and centrally managed (SDN
controllers).

20 May 2016
Technology
Telco Cloud
Figure 20: Internet’s black boxes
App App App
Operating System
Packet forwarding
hardware
App App App Router App App App
Operating System Operating System
Packet forwarding Packet forwarding

hardware hardware
Router Router
Figure 21: Software-defined network (SDN)
App App App
Operating System
Dedicated packet
forwarding hardware
Dedicated packet Dedicated packet

forwarding hardware forwarding hardware
Do you need both NFV and SDN for telco cloud to work??
NFV and SDN are mutually beneficial, highly complementary to eachother, and
share the same feature of promoting innovation, openness and
competitiveness. However, the virtualization of network functions does not rely
on SDN technologies, and vice versa. For example, if we consider a router
service at a customer site: NFV can be applied by virtualizing the router
function, which replaces the router with a simpler device at the customer site.
SDN can be introduced on top of this to separate control and data. Data
packets can then forwarded by an optimized data plane, which the routing
software is moved from an expensive location in a dedicated platform to an
optimized location (a virtual machine on a server).
Figure 22: Comparison of NFV and SDN

NFV SDN
Aim Replace network functions from dedicated appliances Separation of control and data planes to centralize the
with software that can run on generic servers control and programmability of network
Intital target market Telcos Data centers
Target devices Industry standard servers or switches Industry standard servers or switches
Key benefits Increase agility, lower cost and complexity Increase agility, lower cost and complexity
For more detail on NFV and SDN reference architectures, please see page 64.

20 May 2016
Technology
Telco Cloud
Key benefits and barriers to adoption
1. Reduced capex
The operator is able to deploy virtual functions in commercial off-the-shelf
(COTS) hardware. This infrastructure is expected to support virtual functions
from multiple vendors. Functions that are non-real time and reside on the
control plane can then run on scalable hardware wherever capacity is
available. Consolidation from an infrastructure point of view can also increase
the utilization of computing and storage resources, ie when capacity is under-
utilized, the resources manager may re-allocate the resources to other network
functions or to the processing of other workloads (eg OSS/BSS). In addition,
operating network functions on standard hardware allows for interface
connections over virtual “links” to be placed on the same hardware. This
potentially reduces both physical ports and equipment required. Total capex
savings are estimated to be up to 20%, but actual numbers remain to be seen.
2. Reduced opex
The deployment of virtual functions as software using cloud techniques
enables scalable automation. The ability to automate on-boarding, provisioning
and in-service activations can yield significant opex savings. Another area for
operational savings is reduced manual intervention. For example, VoLTE has
several node elements in the network which are required to deploy an end-to-
end solution. In a VoLTE deployment with proprietary hardware, it could
require installation of many racks of independently deployed equipment. With
NFV, the same scenario could require a single blade server. Total opex savings
are estimated to be as high as 35%, but actual numbers remain to be seen.
3. Service agility – new revenue streams

Virtual functions can be efficiently deployed and connected as new services
without installing proprietary hardware network consisting of multiple boxes.
New revenue opportunities may also arise from telcos’ ability to launch new
services quickly. With the new NB-IoT standard foundational to release 13 of
the LTE standards (LTE Advanced Pro, due in 2017/18), we see tier-1 telcos as
well placed to capitalize on broadening new revenue streams within Internet-
of-Things (IoT) and 5G, and in both cases NFV/SDN are likely to be
foundational to their future success in key new revenue opportunities.
Key barriers to deployment

1. Shortage of skills
Telcos need to build up their own IT expertise in order to design their own
network and take control of their own destiny. However, most have ageing
workforces that have grown up with a traditional networking mindset, many of
whom are not ready to be re-trained. And while AT&T can build a large chief
technology officer (CTO) office and drive extensive work within the open
source community, it is clear that tier-2 and tier-3 operators do not have the
necessary knowhow to take on such a complex transition, and if they are
working on NFV/SDN, it is effectively a “science project” at present. This could
separate telcos into the “haves” and “have nots” on a decade-out view.
Furthermore tier-2 operators may struggle if they completely depend on third-
party system integration/consulting capabilities from the network equipment
vendors and/or IT consulting players. The traditional IT vendors (Accenture et
al) lack the domain expertise in telecoms while the network equipment
vendors may not see the business case for virtualization, or may migrate those
operators to “vendor locked in” solutions.

20 May 2016
Technology
Telco Cloud
2. Lack of interoperability
The NFV/SDN architecture depends on adoption of open standards. This
necessitates vendors to support and develop these standards so that
interoperability is not an issue. We have already heard of cases where vendors
have been reluctant to contribute to open standards, or only pay “lip service”
to the concept. This appears most commonly leveled at Ericsson. Overall
standardization is still a work-in-progress, which could slow down change.
3. Integration with legacy equipment

NFV and SDN are both conceptual approaches to networking that harness
many of the benefits that have already been realized in the IT environment.
However, telcos face a more challenging real-time environment, and they have
a colossal amount of legacy equipment, which leads to integration challenges.
Orchestration will be important to create co-ordination between physical and
virtual network functions, while critical support systems like OSS/BSS which
manage billing and quality of service (QoS) will also be challenged to work
with virtual and physical alike.
We summarise a full list of benefits and challenges of NFV and SDN below.
Figure 23: NFV and SDN: Full list of benefits and challenges
NFV SDN
Benefits Benefits
Reduced capex via low-cost COTS servers (x86) Centralized provisioning and network control
Multi-versioning and multi-tenancy of network functions Service provisioning speed and agility
Operational efficiency through single platform and automation Network flexibility
Faster service delivery (time-to-market) Better and more granular security
Service differentiation and customization Reduced complexity, efficiency and lower opex
Reduced opex costs (power consumption, reduced space & monitoring) Reduced capex via reduced dependency on proprietary hardware (eg
lower cost switches)
Allows for multi-vendor interoperability (reduced vendor lock-in) Automation
Challenges Challenges
Lack of standardization and technology maturity Single point of failure
Lack of relevant skills, both operator and vendors Additional bandwidth
Openness and questions over interoperability Set-up delay
Integration of physical (legacy) and virtual infrastructure, and multiple vendors Lack of standards (eg northbound APIs)
Carrier-grade scalability and robustness (hardware accelerators, high availability, Lack of service control software
service level agreement (SLA) support)
Real-time and dynamic provisioning Multi-vendor network control
Management and orchestration (MANO) of virtual appliances
Resistance to loss of vendor lock-in and silo platform issue
Risks to network performance
Hard to measure benefits (TCO uncertainty)
Integration with or replacement of OSS/BSS functionality
Potential benefits to waiting and watching – “fast follower” strategy
We also show the results of a survey of operators from IHS Infonetics. What is
interesting in our view is that agility of service provisioning is seen as the key
benefit of NFV and SDN. On the barriers, availability of talent, integration with
legacy hardware and integration with OSS/BSS as key issues. This illustrates
that while NFV/SDN offer much promise, implementing full NFV and SDN will
be a 10-year transition. Many barriers to adoption exist, and only tier-1
operators appear prepared to take on the challenge given the relative scale of
their CTO offices.

20 May 2016
Technology
Telco Cloud
Figure 24: NFV deployment drivers Figure 25: SDN deployment drivers
Simplification and automation of

Use software for quick revenue 82% 79%
service provisioning
Simplification and automoation of
75%
network provisioning
Increase operational efficiency 77%
Service automation 71%
Scale services up or down quickly 55%
End-to-end service management 61%
New services/efficiences not

55% End-to-end network management 57%
possible with current tech
Quicker introduction of new
Use commercial servers, not 54%
50% services for faster time to revenue
network equipment
Network virtualization 50%
Multi-tenancy 27%
Orchestration over multi-domain,
50%
multi-layer, and/or multi-vendor
Save energy consolidating
27% Network programmability 46%
workloads
Open standards and open

VNFs from small players 23% 39%
softtware to avoid vendor lock-in
Modify services faster 39%

Real-time network optimization 18%
Reduce router/switch cost 25%
Source: IHS Infonetics Source: IHS Infonetics
Figure 26: NFV barriers to adoption Figure 27: SDN barriers to adoption
Integrating SDN into existing

Products not carrier grade 59% 84%
networks
Immature technologies and
Finding/training staff 59% 81%
products
Integrating SDN-based elements
OSS/BSS for NFV 50% 60%
into OSS and back office systems
Incomplete or non-existent
Unknown TCO 50% 43%
standards
Integrating NFV into existing Lack of SDN
45% 39%
networks controller/orchestratoin…
Unknown ROI 45% Unclear or unknown TCO 39%
Lack of open NFV orchestration Deciding which SDN

41% 38%
software controller/orchestration…
Test/service validation too complex Finding/training staff 32%

32%
Unclear or unknown ROI 26%

CIO-CTO responsibility splits 32%
Uncertainty due to competing
26%
Incomplete standards 27% standards initiatives
Seleecting/prioritizing where to
4%
Selecting NFV use cases 14% deploy SDN in the network
Source: IHS Infonetics Source: IHS Infonetics

20 May 2016
Technology
Telco Cloud
Operator engagements to date

Moving from the proof of concept stage into the deployment phase
To date, leading-edge operators are actively involved in proofs of concepts
with ETSI (European Telecommuncations Standards Institute). We also assess
their level of commitment based on public statements. However, starting in
2016, we expect to see many PoCs entering limited deployments, which
should provide the basis for broader deployments throughout the network.
Figure 28: Tier-1 innovator operators actively working on NFV/SDN

Operator Number of proof of concepts Percentage of global wireless
(PoCs) on NFV with ETSI capex
Telefonica 8 4%
AT&T 7 6%
Deutsche Telekom 4 1%
Vodafone 3 5%
NTT 3 3%
BT 3 0%
Telenor 3 0%
SK Telecom 2 1%
Sprint 2 3%
Telecom Italia 2 0%
Verizon 1 6%
China Mobile 1 12%
China Telecom 1 5%
China Unicom 1 3%
Orange 1 2%
KDDI 1 2%
Telstra 1 1%
Sub-total 54%
Source: Deutsche Bank, ETSI
Popular early use cases and timing

In general, early use cases are characterized by high potential ROI, high control
plane intensity, and a locality closer to the core of the network. However,
achieving full virtualization of all parts of the network looks like a 10-year
transition.
In Figure 29, we show the areas of the network that are currently seeing the
early use cases. So far, virtualized customer premise equipment (vCPE) is the
number 1 use case given that vCPE enables today’s complex boxes at
customer sites, whether at business or home, to be replaced with much
simpler boxes while many network functions (eg firewalls, routing) can be
pulled back into the carrier’s network. This leads to cheaper boxes, reduced
support costs, and reduced “truck rolls” - the benefits are compelling. The next
biggest use cases are i) service chaining (connecting L4-L7 network services in
a virtual chain), ii) vNPaaS (virtual network platform as a service), which should
enable scalability and reliability of applications, iii) virtual provider edge (could
be telco OTT services), and iv) virtualized EPC (evolved packet core), which is
the ability to run the mobile core on commercial off-the-shelf servers, rather
than proprietary hardware.
In Figure 30, we show the results of a survey commissioned by Nokia
presented at the Cloud Standard Coordination Workshop in January 2016,
which appears to have more of a sample bias towards mobile operators. Here
virtualizing the EPC comes first as the sample’s priority for virtualization.

20 May 2016
Technology
Telco Cloud
Figure 29: Deployment timing for NFV use cases
2015-2016 2017 or later Won't deploy/don't know
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
vCPE 64% 27% 9%
Service chaining 36% 45% 19%
vNPaaS 32% 45% 23%
Provider edge 27% 50% 23%
vEPC 23% 45% 32%
Consumer home 23% 41% 36%
vIMS 18% 59% 23%
vCDN 18% 27% 55%
GiLAN 14% 41% 45%
vRAN 9% 55% 36%
Consumer fixed access 9% 45% 46%
vBNG 5% 64% 31%
Source: IHS Infonetics
Figure 30: What are your company’s top three priorities for network virtualization?
Evolved Packet Core (EPC) 86%
Virtual private networks (VPN) 84%
Virtual managed services (DPI) 75%
IP applications (eg VoLTE etc) 73%
Voice core (IMS and softswitching) 54%
Network edge gateways, SBC 50%
Content delivery network 48%
Virtualized enterprise CPE (vCPE) 37%
Virtualized consumer CPE (eg home) 32%
Policy control and signaling plane 28%
Radio access network (RNC, cloud RAN) 24%
Online charging (OCS) functions 18%
Source: Nokia, Heavy Reading, August 2015

20 May 2016
Technology
Telco Cloud
Focus on mobile networks

In this section, we focus in on the impact of NFV and SDN on mobile
networks, given our coverage of Nokia and Ericsson. Mobile infrastructure
spending comprises 64% and 82% respectively of the revenue of those
companies, by far the highest exposure to mobile infrastructure of the large
network equipment vendors.
Key points
 The initial focus of “true” virtualization for mobile operators is in the
evolved packet core, which comprises only 4% of mobile capex spend
 Cloud RAN (C-RAN), harnessing cloud concepts, looks set to impactful
in the near-term, rising to 33% of mobile infrastructure spend by 2020
 Operators and vendors disagree on the feasibility of virtualizing the
access network (RAN), but this could be a real threat by 2022
The LTE network architecture today
Over the last 6 years, LTE networks have gradually become pervasive, having
been launched by than half of the worlds 800+ mobile operators. LTE was a
break-through technology at launch in 2010, which created a simplified
infrastructure, and cell sites with fewer pieces of equipment to manage. It also
distributes functions found in 3G core networks to equipment in the cell site.
Figure 31: The architecture of the LTE mobile network

Fiber
Microwave Control and signaling plane
Copper
User or data plane
OSS/BSS
BSC SGSN
BTS
2G Operator IP
HSS PCRF Services
(IMS)
RNC
Node B 3G
MME
UE
SGW PGW Internet

SGi-LAN
eNode B
LTE
65% of cost 15% of cost 4% of cost
Radio access (RAN) Backhaul Evolved packet core (EPC) GiLAN Internet
Source: 3GPP, Deutsche Bank
Notes: UE = user equipment, BSC = base station controller, SGSN = serving GPRS support node, MME = mobility management entity (control plane – session management) , SGW = serving gateway, PGW = packet
data network gateway, HSS = home subscriber server (user identification), PCRF = policy and charging rules function - manages the service policy, OSS = operation support systems (fulfillment, assurance – service
quality); BSS = business support systems (billing, CRM, customer care); Gi-LAN = gateway internet LAN interface

20 May 2016
Technology
Telco Cloud
The LTE network consists of five main parts:
The radio access network (RAN): the RAN connects individual devices to other
parts of the network through radio connections, and accounts for around 65%
of network equipment cost. An LTE cell site consists of antennas, amplifiers to
adjust power levels, and a main piece of equipment, the evolved eNodeB
(eNodeB), aka a base station. The eNodeB allocates radio frequency to devices
and passes calls off to other cell sites and to the mobile operator’s core IP
network. The eNodeB also manages Multiple-Input Multiple-Output (MIMO)
antenna functions and Orthogonal Frequency Division (OFDM) signaling. The
RAN is a key element of capex intensity and operator’s opex (power).
The mobile backhaul: the backhaul is the series of connections from the base
station to the mobile core. This accounts for 15% of network equipment cost.
The means of transport can be fiber cables, microwave or copper wires, with
fiber rising in importance. With LTE, backhaul has moved from TDM to all-IP
(IP/Ethernet). Edge routers now comprise 40-45% of this market.
The evolved packet core (EPC): The evolved packet core (EPC) integrates the
traffic of multiple RANs and is the central point of data plane intelligence. This
accounts for 4% of network equipment cost. The EPC consists of four network
elements: the Serving Gateway (S-GW), the packet data network (PDN)
Gateway (P-GW), the policy and charging rules function (PCRF) and the home
subscriber server (HSS). The S-GW deals with the data plane, transporting data
between the UE (user equipment) and the external networks. The P-GW is the
point of interconnect between the EPC and the external network. The MME
(mobility management entity) deals with the control plane. It handles the
signaling related to mobility and security for access. Finally, the HSS is a
database that contains subscriber-related information. The HSS also provides
support functions in mobility management, call and session setup, user
authentication and access authorization.
Gi-LAN/Internet: The Gi-LAN is the gateway internet LAN interface, and this is
where the network operator provides content optimization solutions as well as
support functions (filtering, DNS, firewalls). The Gi-LAN’s main function is to
handle network flow for network services. Examples of functions include
firewall, network address translation (NAT), load balancing, content delivery
network (CDN), and deep packet inspection (DPI). The Gi-LAN is characterized
by static service chaining, which lowers flexibility and increases costs.
Applications, services: Service support functions like IP Multimedia Core

Network Subsystem (IMS) look set to shift to support open standards (eg
OpenStack, Hadoop). IMS provides the foundation for many services including
VoLTE (voice over LTE).
Virtualization impact on mobile networks
RAN: ongoing migration to cloud RAN (C-RAN), and eventually vRAN

In conventional LTE, the radio access network (RAN) performs both radio and
baseband processing inside a base station. There are two types of C-RAN – the
more basic, centralized RAN and the more difficult to realise, “cloud RAN”.
Confusingly, both are described as “C-RAN”. Centralised RAN has been around
for a few years, and de-couples base station functions into distributed remote
radio heads (RRHs) at the cell site, and centralized baseband unit (BBUs). RRHs

20 May 2016
Technology
Telco Cloud
connect to BBUs via a fronthaul network, which generally needs good

availability of fiber given high data rates over the CPRI interface. This type of C-
RAN is well established in Japan, Korea and China, but the US/Europe offers
the best path for growth for the technology in coming years, although both
regions are stymied by the availability of fiber. One knock-on impact of C-RAN
approaches is an increased scramble around for dark fiber for fronthaul.
Altiostar also claims that its C-RAN implementation can run over microwave
links for shorter ranges, while BT is exploring using G.fast over copper.
IHS Infonetics expects the C-RAN market to reach $16bn by 2020, equivalent
to 33% of the overall wireless infrastructure market (ex services). Nokia is #1
(25% share) in C-RAN given its strong position in Japan. Ericsson is #2 (22%
share), as per Infonetics. We believe Nokia is well placed to grow share here.
The more advanced, yet nascent approach to C-RAN is to virtualize the

baseband functionality completely such that it can be pooled and then
distributed throughout the network as needed. In this way, the baseband
functionality is not centralized in a “baseband hotel”, but instead runs on
common hardware with customized software running on top. This virtualized
RAN (vRAN) approach looks set to materialize from 2022 at the earliest, as
things stand, given severe difficulties around virtualizing layer 1. Some
ambitious startups like ASOCS would though disagree.
Somewhere in the middle between centralized RAN and vRAN lies Nokia’s
recent AirScale C-RAN launch. This product virtualizes layers down to but not
including the time-critical elements of layer 2 and layer 1, which remain local.
We believe this product is market-leading compared with Ericsson, and should
be foundational to 5G. In addition, this product appears to deal with some of
the challenges with the CPRI interface, supporting Ethernet over copper and
fiber. This product should also support local hosting of content (eg video).
Figure 32: C-RAN topology Figure 33: C-RAN market ($bn)
Baseband units (BBUs)

Remote radio heads (RRHs)
C-RAN as a % of mobile infrastructure (ex services)
20 33% 0.40
24%
13% 18%
8% 10% 0.20
15
6.2 0.00
10 4.8
3.7 -0.20
5 3.0
2.2 2.6 9.5 -0.40
5.5 7.2
2.9 3.5 4.3
0 -0.60
2015 2016E 2017E 2018E 2019E 2020E
Source: Intel Source: IHS Infonetics

Definitions:: BBU: The baseband unit processes the baseband signal; the functions include
encoding/decoding, ciphering/deciphering, frequency hopping; RRH: the RF processing unit of a
cellular base station, and is connected to the BBU through a optical or wireless link
China Mobile claims that C-RAN deployments offer capex savings of up to

30% and opex savings of up to 53%. Many of these savings are related to
power and space savings as well as simplification of handover and high
utilization of resources. KT have deployed a commercial cloud-based LTE and
claim a capacity increase of 60-80% because of cell edge improvements.

20 May 2016
Technology
Telco Cloud
Backhaul: SDN to impact

SDN promises to allow network operators to drive network flexibility to drive
down cost (higher resource allocation). According to Infonetics, 45% of
operators expect to deploy SDN in the backhaul network by 2017, while a third
will do so by 2016. The top two reasons for deploying SDN are greater network
flexibility and improved operational efficiencies. A common example would be
using SDN to adjust down bandwidth automatically in microwave links, in
urban/suburban environments where required bandwidth can fluctuate rapidly.
The broader use of femto, pico and small cells would make the use of SDN for
optimization of the RAN backhaul even more compelling.
Evolved packet core: virtualization of the packet core is already happening

Aside from vCPE (virtualized customer premise equipment), which is gaining
traction for enterprise use cases, vEPC appears to us the next most popular
use case for NFV. According to a survey commissioned by Cisco, 46% of
operators surveyed plan to deploy vEPC in the next 3 years. The prime targets
within the packet core for virtualization are the PCRF (policy server) and the P-
GW and S-GW. The key use cases for vEPC are capacity expansion, service
flexibility, IoT, and MVNOs. Virtualizing the EPC functions enables telcos to run
functions as virtual machines running on generic hardware, instead of being
implemented as dedicated equipment. A study commissioned by Affirmed
Networks (a key startup) in July 2015 has estimated that operators can save
68% on capex and 67% on opex on the evolved packet core (EPC) alone.
According to their survey, telcos were also able to deploy a virtualized network
in 6 months versus traditional networks at 15 months on average.
Gi-LAN: virtualizing the Gi-LAN

Virtualizing the Gi-LAN is currently less popular than virtualizing the packet
core, but still attracts high interest. The SDN controller should in theory be able
to programme and dynamically configure network services. This should reduce
the need for manual changes at the hardware level. Virtualizing the Gi-LAN
could then replace the static service chains in place today with dynamic
service chains, interconnected by a virtual switch. The layer 3-7 functions that
operate on top of an IP transport network would be the most likely candidates.
These include “middle box” functions such as firewalls, traffic optimization,
carrier-grade NAT, and content delivery networks (CDN). For example, DPI
(deep packet inspection) can be virtualized to run on commercially off-the-shelf
(COTS) hardware to be deployed pervasively in the network.
Applications, services/OSS/BSS: virtualization ongoing

Virtualization is also applicable to the functions of the IMS (IP Multimedia
Subsystem) which defines and architecture for enabling the delivery of
multimedia services such as VoLTE (voice over LTE). Metaswitch Networks has
already seen significant traction in vIMS, while equipment vendors like Nokia
(Liquid Core) also have presence. Key elements such as SBCs (session border
controllers) are widely available as virtual functions. The key driver is flexibility
to scale as needed. Amdocs has been offering virtualized real-time charging
(BSS) since 2013 (with VMWare), while 66% of operators are planning to
virtualize their BSS offerings by 2016 according to Telecoms.com. Ericsson’s
approach here has been to offer private cloud services (IaaS), either at
Ericsson’s or client’s sites, to support transformations including OSS/BSS,
including Intel hardware. However, we believe that the hardware margin will
be low, despite claims from both Nokia and Ericsson that their Intel-based
hardware has significant customization for “telco grade” requirements.

20 May 2016
Technology
Telco Cloud
We show below in Figure 34 a simplified summary of the pace of virtualization

at mobile operators. As the chart illustrates, the number of cell sites run in a
large country-wide network runs into the hundreds of thousands, which makes
the transition to a virtualized RAN (vRAN) architecture even more challenging.
Figure 34: Summary of the pace of virtualization in mobile networks

Location # sites Pace of virtualization
Large data center 1-10 Fast
Metro data center Up to 200
Central office Up to 1,000
Slower
Cell sites Up to 100,000
Figure 35: General architecture of a SDN and virtualization-based mobile network
Data/User Plane Control Plane Application Plane

Ext. Network
IMS System Internet

SDN-Service SFC QoE/QoS PCEF
vDPI vFW vIDS vDPI vFW vIDS Controllers

Other Network Management
Cloud-based Pool PCSCF IoT
Apps
IMS System
vNAT vLB Proxy vNAT vLB vProxy Plane
Southbound Interface (Openflow, BGP, ForCES, etc. …)
E/W bound
Interfaces
Middleboxes NFVI
Northbound Interface (RPC, JSON, REST, etc. …)
Virtual
Mobile
Mobile Core
Network
MME GW-C HSS PCRF Operator A
GW-D
GW-D SDN-Core
vGW-D vGW-D
Controllers ANDSF ePDG-C AAA
GW-D
Pool Virtual
vGW-D
GW-D Other Network Mobile
E/W bound
Interfaces
GW-D NFVI OCS OFCS

Apps Network
Operator B
vRouter
Switch vCPE
Backhaul
Router
SDN-Backhaul Routing TE VPN Virtual
Mobile
vSwitch Controllers Network
Other Network
Pool Monitoring
Apps
Operator C
Openflow Openflow NFVI

E/W bound
Interfaces
BACKHAUL NETWORK (OPENFLOW-BASED MPLS, OPENFLOW-BASED OPTICAL…)

…
Radio Access
Node B Node B eNode B eNode B RRH RRH

AP AP RRH
3G LTE (4G) Interference Off- D2D/
SDN-RAN Management loading M2M
Virtual
Mobile
Node B
eNode B RRH
RRH Controllers Network
eNode B AP Other Network Operator N
Pool RRM SON
Apps
UE
Source: Van-Giang Nguyen, Truong-Xuan Do, YoungHan Kim, School of Electronic Engineering, Soongsil University, Seoul, South Korea (openaccess: http://link.springer.com/article/10.1007/s11277-015-2997-7)

20 May 2016
Technology
Telco Cloud
In Figure 35, we show a more complicated chart, which summarises how a

virtualized mobile network could look in 2020. This does not factor in any use
of virtualized RAN, but shows the use of virtual routers in backhaul, vIMS, and
a virtualized mobile core.
Operator and vendor views diverge on the pace of change
While operators and vendors are agreed on the need for change, it is
interesting that there is no consensus at vendors or at operators on the likely
pace of virtualization. We show below some example views from Telenor,
Ericsson and Nokia.
I) Operator view – Telenor
Figure 36: Cost-benefit and performance
Source: Telenor; red highlights by Deutsche Bank
Figure 37: Maturity of use cases
Source: Telenor; red highlights by Deutsche Bank

20 May 2016
Technology
Telco Cloud
II) Vendor view – Ericsson
Figure 38: Ericsson’s view on portfolio migration: value versus risk
High Media distribution

network
Hosted
Control plane elements, CSCF, MSC managed
services
Gateways and Appliances
Distributed Cloud
OSS/BSS
Home appliances
Value
Edge router
EMS
Real time
OSS/BSS
Core Home
router networking
Radio Fixed
Access Access
High Low
Risk
(Technology maturity, performance requirements)
Source: Ericsson, “Communications as a cloud service”, July 2014” red highlights by Deutsche Bank
I) Vendor view – Nokia
Figure 39: Nokia (ALU)’s view on virtualization benefits: automation gains vs cost gains
High
MEDIA SERVERS
AND CDN
CPE
NETWORK
MEDIA GATEWAYS APPLIANCES
Automation gain
NETWORK
APPS
PACKET GATEWAYS ANALYTICS
CONTROL PLANE CHARGING PLATFORM
Radio access FUNCTIONS AAPLICATIONS
networks NFV/SDN
EMS/NMS CONTROL
Edge Automation and Cost (TCO)
OSS/BSS
Routers optimization • Capex
Core • Elastic scale • Opex
Routers and • Resource Pooling
Ethernet • Rapid deployment
switches • Location optimization
Low
Less More
Cost gain
Source: Alcatel-Lucent (2013): “Network functions virtualization – challenges and solutions”; red highlights by Deutsche Bank

20 May 2016
Technology
Telco Cloud
We note that operators tend to be more optimistic on the prospects for

virtualization, eg Telenor believes vRAN/vBS (virtual base station) could be
mature by 2018, while Ericsson sees this as “high risk” and “low value”, while
Alcatel-Lucent (now Nokia) initially saw no path to a virtualized router in its
roadmap - Nokia has since launched the Virtualized Service Router for edge
routing, with positive reviews from independent tests by the EANTC.
The ongoing theme is that mobile operators are optimistic about the cost and
agility benefits to their networks, while vendors are treading carefully as they
consider the effects of cannibalization on their own businesses. This theme of
operator/vendor mis-alignment is a key theme that is becoming more
prevalent. Perhaps the best illustration of this challenge has been the lead role
that AT&T has been taking at driving virtualization forwards. Verizon too
recently published a NFV/SDN reference architecture, which appears to be
another attempt by operators to drive the conversation and for vendors to
follow, not the other way around.

20 May 2016
Technology
Telco Cloud
Implications for vendors

Key points
 Large operators like AT&T are increasingly taking matters into their
own hands, to drive transformation in their networks, frustrated with
vendors’ lack of openness as they worry about loss of “lock in”
 The disaggregation of software should lead to rising software sales,
but this may not offset lost profit from lower hardware sales
 Increased competition looks likely across software and services
 Vendors are re-skilling their workforces towards cloud versus the old
network skillset. This should lead to more restructuring in years ahead.
 Telco cloud presents the opportunity of transformation deals, but this
is offset by increased competition from traditional IT services
companies and the prospect of a decline in managed services sales
 The enterprise channel is set to rise in importance for both telcos and
vendors given virtualization enables “network slicing”, a potential
revenue driver for telcos if net neutrality concerns can be allayed
 The financial impact of NFV/SDN on vendors is small initially as it is
confined to contained domains, but this could become large, rising to
54% of equipment spend by 2030.
 C-RAN promises to be more impactful on the wireless equipment
vendors in the near-term, given that it addresses the 65% of mobile
operator equipment spend that is deployed in the RAN.
1) Reduced vendor lock-in – vendors and operators at odds
With considerable capex savings on offer (combined with opex savings and
faster agility), it is no wonder that telco operators are embracing NFV/SDN,
most notably the largest tier-1 operators like AT&T, which is targeting to
virtualize 75% of its network functions by 2020, while Vodafone is looking to
virtualize 50% on the same timeline (announced: May 17th 2016). However, we
should recognise that the technology is still in a “trial phase”, and tangible
savings on trials have been hard to measure. Operators have been struggling
with many issues including poor interoperability, lack of skills, but the most
common complaint we have heard is that vendors are offering “lip service” to
NFV/SDN, but in fact trying to slow down the transition to protect their core
businesses. So the saying goes, “turkeys don’t vote for Christmas”.
Vendors fear the loss of vendor lock-in and the most common area where we
hear complaints around virtualized solutions from the equipment vendors is
around resistance to open standards. Many large vendors are just too resistant
to the idea of “letting go” of a proprietary attitude, and thus are slow to
support open standards like OpenStack, a key pillar of telco cloud. So the
claims of “openness” are often seen as a “proprietary trap”. However, in
addition, sales forces at the big vendors know a negative disruption when they
see one and may not be exactly keen to sell the benefits of a virtualized
solution over a traditional solution especially given the significant and
profitable maintenance contract that is attached to the hardware sale.

20 May 2016
Technology
Telco Cloud
The notable point on the operator side is the level of determination being
shown by operators to make this transformation a reality, and clearly AT&T is
leading the way here, while Verizon, Telefonica, DT, NTT, Verizon and
Vodafone are not far behind.
2) Software unlikely to mitigate lost profit entirely
The move to telco cloud causes significant disruption to traditional network

equipment vendors, as it de-couples hardware and software. Until now,
equipment vendors have preferred to sell equipment as an integrated device,
essentially a closed black box. They have then managed the pace of upgrades,
with upgrades often the highest margin sale within the customer lifecycle.
With NFV/SDN, the operating system can be supplied by a third party or by an
open source project that provides a network operating system, while hardware
is provided by COTS servers (likely x86).
Legacy hardware sales likely to decline with software growing

With telco cloud set to take off, dedicated hardware sales are likely to decline
significantly over the next decade. The good news is that the rising software
revenue opportunity should help to mitigate the lost profit pool from declining
hardware sales. However, we don’t believe that the profit pool growth in
software can completely offset the lost profit in hardware. This is similar to our
view of the wireless infrastructure market today, where we see software
upgrades (to LTE-Advanced and LTE-Advanced Pro) rising to help mitigate a
fast declining hardware market.
3) Increased competition from IT vendors and startups
We show in Figure 40 the new “lie of the land” in telco cloud. What can be
seen is that existing network equipment providers are now competing with an
increased number of stakeholders.
The re-emergence of IT vendors

The first group that equipment vendors need to face off to is traditional IT
hardware companies. In response to their more limited track record in cloud
computing, the likes of Ericsson and Nokia have been building their own data
center capabilities to support telcos to develop private cloud networks. This
has also involved launching Intel-based hardware (Ericsson’s HDS 8000, RSA-
based, Nokia AirFrame) to not allow this business to naturally migrate to the
classic x86 server vendors like Dell and HP Enterprise. HP in turn has been
active in building its telecom domain expertise, acquiring ConteXtream in
2015. To date, HP’s C7000 blade servers seem the most popular hardware
underlying telco cloud deployments, while Dell is only really just beginning to
attack this vertical.

20 May 2016
Technology
Telco Cloud
Figure 40: Telco cloud: networking and IT worlds collide
TELCO Dell, IT
Brocade
Ericsson, Nokia, IT WEB 2.0
Cisco, Huawei, hardware
Ciena, Juniper, AWS
Samsung, ZTE vendors
Network IT
equipment consulting
providers firms
Telco
Cloud
HP Enterprise,
Accenture
Amdocs,
Metaswitch
Networks, Affirmed Networks,
Oracle, Domain Altiostar, Cumulus,
NEC Startups Big Switch
experts
Networks, Pica8,
Pluribus, ASOCS,
6Wind
In terms of IT services vendors, telco operators have been courting the likes of
HP and Accenture to get involved in cloud transformation projects. The
attraction is their cloud expertise and relative neutrality compared with
equipment vendors. However, so far this strategy has back-fired with
Telefonica removing HP as a lead partner its Unica project. To date, their
telecom domain expertise is insufficient.
The emergence of startups or smaller companies

Operators have been actively seeking out startups and small companies to help
drive the transformation. With open standards and disaggregation of hardware
and software, startups should be in a position to flourish. AT&T has signed up
Affirmed Networks for its vEPC and Metaswitch Networks for its vIMS. These
startups are unencumbered by legacy businesses and generally offer very
attractive pricing and new business models (eg subscription-based). The flip
side is that these companies are small and they may need operators to help
them to scale up their businesses and make them financially viable. We could
then see more venture capital investment from operators ahead from the likes
of AT&T Ventures, Vodafone Ventures, Verizon Ventures, T-Ventures et al.
4) Services – a mixed picture
We believe that cloud disruptions lead to tier-1 operators becoming more

selective about vendors, including their system integrators. While the positive
implication of cloud is that Ericsson and Nokia get involved with more and
more “transformations” and they have more “domain knowledge”, we believe
the prospect of these vendors securing a lock on that operator are likely to
diminish with cloud, not helped by new entrants like HPE and Accenture. In
our view, this challenge will echo the challenges experienced by IT outsourcing
vendors in the IT cloud disruption (eg IBM, CSC et al). We also see network
rollout and managed services as being impacted on a decade out view. So the
net impact could be negative, especially given a potential skills mis-match.

20 May 2016
Technology
Telco Cloud
We expect Ericsson to be most vulnerable to this trend given its 66,000

services personnel, while we also expect cloud implementations over time to
lead to a reduced dollar opportunity for managed services. For Ericsson in
particular, we have also already seen in Q1-16 how OSS/BSS transformations
carry significant risk in early stages given scoping issues. Ericsson’s systems
professional services margin fell from 12% to 7% in 12 months, driven by a
handful of major transformation deals. We also see Ericsson’s high exposure
to managed services and network rollout (NRO) as a key risk as we move into
the next decade given that both these segments could see an impact from
increased adoption of telco cloud.
Figure 41: Telecom services market ($bn)
350
300 Operator internal spend
250
Network rollout (NRO)
200
150 Managed services
100
Consulting, systems
50
integration & customer
support
0
2011 2012 2013 2014 2018E
Source: Ericsson Capital Markets Day, November 2015
We show in Figure 42 the headcount of Nokia and Ericsson, and how Ericsson
is lop-sided in its headcount towards services. In Figure 43, we show how
Nokia and Ericsson are the most exposed to this trend – much more than
Huawei, Cisco, Juniper and Ciena.
Figure 42: Headcount by geography and function – Nokia larger in R&D, Ericsson larger in services
Nokia Alcatel-Lucent New Nokia Ericsson
Europe 22,171 17,000 39,171 45,602
Asia Pacific 25,148 16,000 41,148 39,759
North America 3,665 12,500 16,165 14,548
ROW 4,734 4,500 9,234 16,372
Total headcount 55,718 50,000 105,718 116,281
Of which R&D 20,000 20,000 40,000* 25,700
Of which services 40,000* 66,000
Of which other (manufacturing, SG&A) 25,718 24,581
Productivity metrics
Revenue per employee (€), 2016E 215,033 283,241 245,438 220,984
R&D spend per R&D employee (€), 2016E 107,154 118,725 112,940 130,247
Source: Company statements, Deutsche Bank estimates *both are likely to reduce to 35-36k given ALU synergy 33ealization

20 May 2016
Technology
Telco Cloud
Figure 43: Telco services as a % of sales
Telco services as a % of total sales
42%
32%
17% 18%
4% 3%
Ericsson Nokia Huawei Juniper Cisco Ciena
5) Organizational disruption is ongoing
With such disruption ahead, it is no surprise that equipment vendors like

Ericsson, Nokia and Cisco have been busy transforming their own businesses
in response to the emerging transformation at their tier-1 customers, who
typically represent 50-55% of sales. We show recent organizational events at
Cisco and Ericsson in Figure 44.
In general, the trend is to create a dedicated cloud hardware and software

segment, but Ericsson has recently gone further and merged its cloud services
unit into an enlarged cloud group. While it makes sense to create a dedicated
group focused on cloud, there is a flip side, namely that the traditional
hardware segment become disconnected and threatened by the separation.
For example, we have met startups that have ridiculed the separation between
Ericsson’s radio and cloud units as working against eachother, rather than with
eachother.
Figure 44: Recent organizational events at Cisco and Ericsson

Date Company Reorganization/partnership
Apr-16 Ericsson Ericsson forms two new units - IT & Cloud Products, and IT & Cloud Services. Services split into Network and Cloud Services
Mar-16 Cisco Engineering to be reorganized into four units: Networking, Cloud Services & Platforms, Security, Applications & IoT
Nov-15 Ericsson/Cisco Ericsson and Cisco announce strategic partnership to drive innovation and speed digital transformation across industries
Oct-14 Cisco Engineering moving from teams being focused on products to one group focused on software, and another on hardware
Apr-14 Ericsson Ericsson splits Networks division into Radio and Cloud & IP
The hunt for skills – often via M&A

We show in Figure 45 overleaf the key M&A to date in this space. Vendors
have been scrambling not only to find the necessary skills, but also to re-
organize internally. Most notably, Nokia’s acquisition of Alcatel-Lucent was
tied to the belief that a full broad-based understanding of the whole telecom
network (optics, routing, access) is now necessary in an NFV/SDN world while
customers have also moved towards fixed-mobile convergence (FMC). We are
aligned with that view. Ericsson’s strategic partnership with Cisco in

20 May 2016
Technology
Telco Cloud
November 2015 has given it undoubted product strength, but we continue to

have doubts around the feasibility of the alliance, given the confused roadmap
being presenting to operators with some products competing with each other
in areas like mobile core and IP routing.
Figure 45: Notable M&A in telco cloud / next-generation networking

Date Acquiror Target Valuation ()USDbn) Description
Apr-16 Brocade Ruckus 1.2 WiFi networking
Apr-16 Ericsson NodePrime - Data center management
Mar-16 Cisco Jasper 1.4 Internet of things (IoT)
Feb-16 Nokia Nakina Systems - Security and orchestration software
Feb-16 Mellanox EZchip 0.8 Network processors
Jan-16 Nokia Alcatel-Lucent 16.6 Optics, routing, wireless infrastructure, NFV/SDN leadership
Jan-16 ADVA Overture 0.0 NFV/SDN - network orchestration
Dec-15 Intel Altera 16.7 Leading FPGA chip provider
Aug-15 Ciena Cyan 0.4 NFV/SDN - network orchestration
Jul-15 Huawei Amartus - SDN
May-15 Nokia Eden Rock - Self-optimizing networks (SON) software
May-15 HP Enterprise ConteXtream - NFV/SDN - network orchestration (service chaining)
Apr-15 Cisco Insieme - SDN
Mar-15 Brocade Connectem - NFV/SDN - vEPC
Oct-14 Ericsson Sentilla - Data center infrastructure monitoring
Sep-14 Ericsson Apcera - PaaS
Jun-14 Red Hat eNovance 0.1 Open source cloud services
Jun-14 Cisco Tail-f 0.2 NFV/SDN - network orchestration
Mar-14 Cisco Altiostar investment Virtualized RAN
Jan-14 Oracle Corente - SD-WAN
Nov-13 Amdocs Celcite 0.1 Self-optimizing networks (SON) software
Nov-13 Cisco Insieme Networks 0.9 Hardware-based response to NFV/SDN
Oct-13 NTT Com Virtela 0.5 SDN, Network optimization, enterprise cloud
Sep-13 Amdocs Actix 0.1 Network performance optimization
Feb-13 F5 Networks Line Rate Systems - SDN for data centers
Feb-13 Oracle Acme Packet 2.1 Session border controller (SBC), IMS diameter signaling
Feb-13 Cisco Intucell - Self-optimizing networks (SON) software
Dec-12 Juniper Contrail 0.2 SDN
Nov-12 Brocade Vyatta - Virtual routing
Jul-12 VMWare Nicira 1.3 NFV/SDN pioneer
Other notables
Jan-12 Ericsson Telcordia 1.2 OSS/BSS services
Dec-10 Ericsson Optimi Network optimization software
Dec-09 Cisco Starent 2.9 Mobile core
Restructuring is becoming a necessity

The main impact though is on the need for restructuring as many employees
are close to retirement or are in no mood to learn cloud concepts having spent
20-25 years working in a classic networking environment. It is this challenge
that will lead to an ongoing need for restructuring in coming years at the likes
of Nokia and Ericsson. We continue to believe that Nokia’s reported plans to
reduce headcount by 10-15k (up to 14% of headcount) are necessary as a way
of offsetting declining end-markets in the core radio hardware market. It is
clear that Ericsson needs to take further action than the steps it has already
taken, and we fear that delays on that front will impact both the share price
and Ericsson’s ability to fund expansion in cloud efforts in coming years.

20 May 2016
Technology
Telco Cloud
A leap to diversify remains a risk

In the face of a slowing telco market, Ericsson’s approach has partly been to
explore new growth opportunities outside of telco. Most notably this applies to
Ericsson’s TV and Media business, which on Ericsson’s own estimates has
around 5% share of a fragmented $15bn market. While there is some merit to
building capabilities around content delivery given the movement of content to
the edge of the network, we remain concerned that Ericsson could do another
deal here as a means of diversifying its business. This also applies to Nokia,
but here we continue to worry that Nokia believes it has a strong future as a
B2C company. This was best illustrated by the recent Withings acquisition.
6) Enterprise channel is growing in importance
Alcatel-Lucent’s success with Nuage in SDN (a spin-in formed in 2011) was

directly related to the fact that the unit was allowed to operate independently
and focus on the cloud transformation. However, another key ingredient was
that the startup entity was focused initially on the enterprise channel (example
customers: Grupo Santander). This enabled them to gain expertise with SDN in
advance of wide adoption of the technology by telcos. The recent launch of the
Network Services Platform (NSP) in 2015 was the fruit of that learning, and is
integrated with CloudBand for NFV orchestration.
One of the key motives of Ericsson partnering with Cisco was to access the
enterprise channel. Ericsson is thinking long-term here, and this very much tied
to the emergence of internet of things (IoT) and 5G. We believe network
equipment providers need to build a broader understanding of the different
requirements of verticals outside of the traditional telco domain. This is
because telco operators are themselves looking at big opportunities to grow
revenue outside of traditional avenues, but with key focus on certain use
cases: examples include advanced manufacturing (industrial 4.0),
transportation (autonomous driving), utilities (smart grid), government (public
safety). Ericsson calls this the “networked society”. While this has been
somewhat over-hyped in recent years, in our view, this will become
increasingly important from 2020 onwards.
Virtualization is foundational to 5G and IoT

Given the slow pace of standardization efforts, we continue to be skeptical on
meaningful 5G deployments before 2021. However, it is clear that a new radio
access standard (including use of new mmWave spectrum for ultra-dense
deployments) will not be the key driver of 5G adoption; rather it will be use
cases that will develop over time for specific verticals. These use cases should
be the key reason to migrate to 5G, while much of LTE equipment and
spectrum will be re-used in many instances, including using NB-IoT in the
guard band. Implementing these uses cases efficiently will depend heavily on
“network slicing”. Network slicing is the creation of a logical network, which is
customized via software for a specific application. And it is virtualization that
substantially lowers barriers to entry to deploying network slicing, via
increased flexibility and decreased cost. This could include content-centric
scenarios.
We note that Telefonica expects the boundary between core and access
networks to blur in 5G networks. In some deployment scenarios, radio
interface functionalities will be centralized in data centers. In others, local
breakout will happen very close to the radio elements, so most of the core

20 May 2016
Technology
Telco Cloud
network functions will be located close to the antennas. The 5G architectural

framework should allow for the support of both extreme cases, as well as
other intermediate alternatives: different levels of radio interface centralization
(in aggregation point), while network intelligence moves closer to the edge (eg
content caching).
The importance of a broader regulatory view on net neutrality

As mentioned, one of the key business cases for investing in 5G and NFV is
“network slicing”. This opens up questions around net neutrality. The latest EU
regulations on net neutrality offer scope for providing “specialized services”,
but there is some disagreement in the industry over what this means in
practice. As reported by Light Reading, Deutsche Telekom believes the
legislation could allow it to charge OTT players for an assured service level,
while Telenor is more skeptical. Telenor is working towards two scenarios of
the world: one when 5G services can be differentiated on latency and speed.
That would allow for price discrimination on different classes of service that
are tailored to vertical and local requirements. However, the second scenario is
that regulation is a barrier to such moves – if that happens, cost reduction will
become more important to Telenor. Such questions around net neutrality could
be a potential “banana skin” for operators to deal with, but we ultimately
doubt that regulation would get in the way of deployment of an ultimately
better technology. However, they may add significant regulatory red-tape in
order to govern pricing.
7) Financials: Impact initially small, but becomes larger
As can be seen in Figure 46, network equipment spending represents about

half of overall telecom capex. Within this, wireless infrastructure on average
represents 18% of telecom capex, but 40% for dedicated wireless operators.
Of equipment spend only, wireless infrastructure represents 37% of spend and
70% for dedicated wireless operators.
Figure 46: Worldwide telecom capex ($337bn), 2015
Wireless
infrastructure,
18% Broadband
infrastructure
Non-telecom
equipment Routers
(labour, fiber,
copper, real Optical
estate etc)
Voice Video
Customer
premises
equipment (CPE)
Source: IHS Infonetics; includes cable operators
We show below industry forecasts on the potential of NFV/SDN spending at

service providers. As can be seen, growth is set to be considerable between
today and 2030. However, the spending is only 13% of today’s spend level by
2020. However, by 2030, NFV/SDN spending is set to represent 54% of
network spending, generating significant cost and capex savings potential.

20 May 2016
Technology
Telco Cloud
Figure 47: NFV/SDN spending, US$bn, service providers
54%
120 44% 49%
% of WW network equipment spend
USDbn
33% 37% 40% 50%

100 27%
15% 19% 22% 87 30%
8% 13% 79
80 2% 2% 3% 5% 72 10%
65
59 -10%
60 54
44
36 -30%
40 30
25 -50%
21
20 13
6 9 -70%
2 4
0 -90%
Source: SNS Research, 2015, Deutsche Bank
Impacted weighted towards mobile core and CPE until 2020

In common with our discussion earlier, between now and 2020, the bulk of
NFV/SDN spending at telcos will be on control-plane centric functions, for
example i) the mobile core (EPC), ii) virtual CPE (vCPE), and iii) virtual IMS
(vIMS). We note that OSS/BSS spending is already largely run purely in
software in an x86 environment, but we include that here as there are still
many telcos yet to transition into a cloud-only environment for OSS/BSS. In
total, the above only represents 10% on the revenue line of the impact of true
virtualized offerings.
At the same time, the architectural shift to C-RAN deployments will likely have
the bigger impact on RAN infrastructure sales given that vRAN is not ready.
However, the ongoing deployment cycle of LTE will be more meaningful for
revenue trends between now and 2018. After that, we see C-RAN (along with
5G) beginning to become as meaningful for the market outlook, with C-RAN
comprising 33% of new RAN deployments by 2020. This is especially true
given that 5G will likely re-use a considerable amount of LTE equipment, while
opening up new mmWave spectrum for ultra-dense deployments, limiting the
hardware opportunity. Overall then, disruptions out to 2020 look set to start
small, but get larger over time.
However, beyond 2020, the largest growth outside of the domains mentioned
above comes from deploying SDN in telco data centers and backhaul
environments, and from virtualizing the radio access network (RAN). This could
then become much more important for network equipment vendors.

20 May 2016
Technology
Telco Cloud
Figure 48: NFV/SDN spending mix (%) by functional area, service providers
Data Center RAN Transport & Backhaul

IMS & VoLTE EPC/Mobile Core CPE
Policy, OSS & BSS Fixed Access CDN
100%
80%
60%
40%
20%
0%
Source: SNS Research
However, we should note that many of these functions are more demanding in
terms of real-time requirements (low latency) and the importance of the data
plane, so this is not a certainty to happen, despite the efforts of Intel and
others to drive this transition. And even by 2030, there will be traditional RAN
sales given that many tier-3 telcos will lag on the transition to a virtualized
RAN environment, given significant incumbency of legacy networks.
NFV software can mitigate declining hardware sales, but not entirely
As can be seen in Figure 49, the NFV market opportunity is likely to be larger
at service providers than SDN for the next decade. This reflects early adoption
of NFV in contained domains at telcos as outlined above. However, as telcos
begin to adopt leading-edge practices in their own data centers, SDN should
catch up by 2030.
Figure 49: NFV/SDN – software mix (%), service provider only
SDN revenue (LHS) NFV revenue (LHS)

SDN software mix (RHS) NFV software mix (RHS)
100 100%
USDbn
90
80 80%
70
60 60%
50
40 40%
30
20 20%
10
0 0%
2015E 2017E 2019E 2021E 2023E 2025E 2027E 2029E
Source: SNS Research

20 May 2016
Technology
Telco Cloud
The software mix within NFV is likely to be high at ~80%, so the profit
opportunity should be significant. However, this move is also likely to be
accompanied by a move to a recurring, subscription-based software revenue
model (eg similar to SaaS companies), as we believe new startups entering this
NFV/SDN domain are pursuing a SaaS-type pricing strategy, and these
companies are unencumbered by legacy business. This will likely mean a net
revenue and profit loss from virtualization over time in our view, especially if
you consider the impact of lost revenue in managed services, which should
decline with easier-to-manage networks.
We note that NFV software is expected to average ~80% of the revenue

opportunity while the data center migration to SDN necessitates significant
hardware upgrades, which keeps the software mix in the 40-60% range.
C-RAN likely to be more meaningful in mobile networks until 2020

Given our focus in this note on Nokia and Ericsson, we outline our overall view
on the addressable market in mobile networks. Overall, we see a gradual
impact, starting in the mobile packet core, while we see C-RAN being
impactful between now and 2020 and leading to share shifts. Beyond 2020,
we see services being impacted from declining managed services sales, and
potentially the introduction of virtualized RAN from 2022. While the software
mix should rise, it will be tough in our view to mitigate the impact entirely.
Figure 50: DB’s view on the pace of adoption of NFV in mobile networks
Adoption curve
2014-15 2016-17 2018-19 2020-23 2024-25
Proof of Early Early Broad Laggards

concepts adopters majority adoption
Key domains Mobile Core (vEPC, vIMS etc), C-RAN vRAN
20%
 Uncertainty around roadmap  vRAN
Net revenue +  Virtualized mobile core  Subscription-model
impact on
 C-RAN savings led by startups
wireless 0%
infrastructure  Managed
market including -  Consulting and SI uplift services decline
services ($100bn)
 White-label Intel hardware sales
-20%

20 May 2016
Technology
Telco Cloud
Underlying wireless equipment market: declining sales, flattish profit

For completeness, we show below our outlook for wireless capex (-7% in
2016E and -5% in 2017E). The key point we would highlight is that software
upgrades are rising in the mix and could be helped by a fast migration to LTE-
Advanced Pro, which should allow the wireless equipment vendors to hold
gross profit steady, even in the face of the tough environment for wireless
capex which is not being aided by the fading of the LTE rollout boom. The
chart in Figure 52 is the point of reference point with which to reference the
red line in the chart above.
Figure 51: Wireless capex outlook ($bn) Figure 52: Wireless infrastructure & services mkts ($bn)
North America Europe APAC LatAm MEA Profit pool Hardware Software Services
180
160 120
100 101 99
140 96 94 93
100
120
80 40 38 40
100 41 42 43
80 60
13 15
60 17 19
40 20 22
40
20 20 47 48 41 36 31 28
0
0
2011 2012 2013 2014 2015 2016E 2017E
2014 2015 2016E 2017E 2018E 2019E
Source: Deutsche Bank Source: Deutsche Bank, IHS Infonetics

20 May 2016
Technology
Telco Cloud
Implications for coverage

Key points
 Six vendors stand out at present as leaders in NFV and SDN – Cisco,
Nokia, Ericsson, Huawei, Juniper and HP Enterprise
 Nokia is the most pervasive vendor, while Cisco has the most
functions virtualized
 Nokia remains our top pick in European telecom equipment
Six vendors stand out at the moment

While the telco cloud changes are in their infancy today to suggest that any
one vendor is “streets ahead” of the competition, we would argue that Cisco,
Ericsson, Nokia, Huawei, Juniper and Hewlett Packard Enterprise are the top 6
players at present, with VMware influential on the hypervisor side (although
deployments appear to be moving to KVM). It remains to be seen how the
Cisco/Ericsson strategic partnership changes this assessment, but we remain
skeptical on the long-term viability of this partnership. The survey of operators
below supports our view of the top 6 vendors, with Cisco top scoring in NFV,
and Nokia (Nuage Networks) top scoring in SDN.
Figure 53: NFV ecosystem – telcos’ level of familiarity Figure 54: NFV vendors – under evaluation
Cisco 64% Cisco 50%

HP Enterprise 55%
Ericsson 36%
Juniper 50%
Nokia 50% HP Enterprise 32%
VMWare 45% Huawei 27%
Huawei 45%
Nokia 23%
Ciena 36%
Ericsson 32% Juniper 23%
Brocade 32% Brocade 23%
NEC 27%
Ciena 14%
Oracle 18%
Amdocs 9% NEC 14%
Figure 55: SDN ecosystem – telcos’ level of familiarity Figure 56: SDN vendors – under evaluation
Nokia 73% Nokia 50%

Cisco 59% Cisco 45%
Juniper 59%
Huawei 23%
Ciena 36%
Ericsson 32% Juniper 18%
Brocade 32% Ciena 18%
Huawei 32% Ericsson 14%
NEC 14% Brocade 9%
Infinera 9%
IBM 9%
ZTE 9%
Fujitsu 9% ZTE 9%
IBM 5% HP Enterprise 5%
Source: IHS Infonetics, June 2015 Source: IHS Infonetics, June 2015

20 May 2016
Technology
Telco Cloud
Public announcements to date

We have also reviewed public statements around virtualization at the key
telcos to understand who is succeeding at which operators. Clearly, being part
of the most important transformations (AT&T, Verizon, Vodafone, NTT,
Deutsche Telekom, Telefonica, SK Telecom) is vital to build expertise in this
early-stage market. For now, most of these large operators are working with
almost all vendors in some capacity, but there should be some consolidation of
vendors over time in our view.
Figure 57: NFV operator announcements to date

Operator Areas of focus to date Deployment Timeline Public vendor(s)
USA
AT&T (Domain 2.0) SD-WAN, vCPE, vEPC, vIMS To virtualize 75% of network by 2020 Ericsson (prime integrator), Cisco, Metaswitch,
(from 5% in 2015). vRAN to come at Affirmed Networks, Nokia, Juniper, Amdocs
the back-end of transformation
Verizon SD-WAN Sep-2015 (initial launch) Cisco, Ericsson, Juniper, Nokia, Dell, Big Switch
Networks, Red Hat, Versa (investor)
Korea
SK Telecom vEPC, vIMS August-2015 (initial launch). Samsung, Ericsson
KT Corp vEPC Large NFV/SDN deployments in 2016 Nokia
LG Uplus vEPC NFV deployment: 2017-2018 Affirmed Networks, Ericsson
Japan
NTT DoCoMo vEPC, vIMS To virtualize 75% of its network by NEC, Fujitsu, Ericsson, Cisco, Nokia
2020. They will virtualize EPC by Mar-
2016 and RAN later
KDDI vEPC (initial trials) 2016-2020, interest in vRAN HPE, Cisco, Nokia, NEC
Softbank vEPC Initial contract in Oct-2015 Cisco
China
China Mobile vEPC, vIMS, vRAN July-2015 (initial launch) Huawei, Nokia
China Unicom vEPC, vIMS June-2014 (initial launch) ZTE, HPE, Huawei, Nokia
China Telecom vCPE April-2014 (initial launch) Huawei, Cisco, Nokia
Europe
Deutsche Telekom SD-WAN, vCPE, vEPC, vIMS, vCDN Cisco, Huawei, Nokia
BT SD-WAN, vCPE Cisco, Huawei, Nokia, Intel
Vodafone vVoLTE, vEPC, vIMS July-2015(initial launch) Affirmed Networks, Huawei, Ericsson
Telefónica vEPC, vIMS, vCPE, vRAN 65% of network to be NFV-compliant Ericsson (prime integrator), Nokia, HPE, Huawei,
by end of 2018 Intel, Broadsoft, Juniper, Brocade, Ciena, Red Hat
Naka vEPC Cisco
Orange vEPC Feb-2015 (initial launch) Cisco, Nokia
Swisscom vEPC, vIMS Ericsson, HPE
Tele2 vEPC Q3-2016 Huawei, Cisco, Canonical
Telia Sonera vIMS, vCPE, vEPC, vSBC Virtualize parts of RAN by 2020 Ericsson, Cisco, Nokia, Metaswitch
Telenor vEPC HPE
Telekom Austria vEPC, vIMS vEPC (Trial in 2014, wide deployment Brocade, Metaswitch, Affirmed Networks
TBA)
Others
Telstra vEPC, vIMS Ericsson, Ciena, Nokia, F5 Networks, HPE
Etisalat vEPC, vCDN, vIMS in 2015 Nokia
Mobily vRAN 2017 Nokia
SingTel SD-WAN, vEPC Deployments between 2016-2017 Brocade
XL Axiata vEPC in 2015 Cisco

20 May 2016
Technology
Telco Cloud
Some initial conclusions

 Nokia is the most pervasive vendor: By mentions: Nokia: 56%, Cisco:
44%, Ericsson: 33%, Huawei: 30%, HPE: 22%. This mainly reflects the
benefits of the combination with Alcatel-Lucent, which was early to
the space, and which brought early relationships with Verizon,
Deutsche Telekom, Telefonica, and in China (China Telecom, China
Mobile). Key solutions include the NFV platform Cloudband and the
Nuage SDN controller, and its cloud optimized OSS layer.
 Ericsson is a little behind in cloud, but leads in services: At the most
strategic operators like AT&T and Telefonica, Ericsson is taking the
lead system integrator role. This could prove to be a strategic position
if it enables them to influence those operators to implement
virtualization according to their own roadmap. However, our own
research and public announcements suggests it is behind Nokia on
NFV/SDN as of today, while there is resistance to going truly “open”.
 Cisco is punching “above its weight”: Cisco gained a foothold in the
mobile core with the Starent acquisition in 2009, but its tradition has
been in selling to enterprises, while it lacks RAN expertise or the
breadth of telco relationships of an Ericsson or a Nokia. Top
management have admitted they were late to SDN, but they claim to
be ahead of the pack on NFV. Cisco had an impressive 49 VNFs
completed by Mar’15, with its next competitor at 10. Cisco claims
that 7 of the largest 10 network providers are customers of Tail-f
(network orchestration).
 Juniper appears focused on the larger vendors: Juniper appears to be
involved with tier-1 operators principally, which may reflect its
reduced reach at traditional telcos compared with its peers.
 HP Enterprise seems well positioned in servers: Hewlett Packard
Enterprise seems to be in pole position to win growing business in
networking via server sales. This is being supported by its stronger
relationships in telcos versus Lenovo and Dell. We note that HPE’s
networking revenue grew double-digits in its last reported quarter
(+54% with the Aruba acquisition). However, it needs more domain
expertise if it is to challenge the likes of Ericsson in services.
 Specialist vendors still play important roles: Specialist vendors
continue to play an important role: we believe that Amdocs for
example recently won a significant role at AT&T in orchestration
where it beat out all of the main vendors. Given that orchestration will
likely be very strategic in full-blown NFV/SDN roll-outs, this is notable.
Implications for the European network equipment universe

 Nokia (Buy, €6.2 target): While ongoing achievement of the ALU deal
synergies will be the #1 stock driver for the next few years, we note
that Nokia looks to be in a stronger position than Ericsson to drive
disruption with the advent of telco cloud. Nokia will likely have 35,000
R&D professionals post restructuring (vs Ericsson’s 26,000), while it
has a long heritage in broad use of open standards and telco cloud
disruptions acquired via Alcatel-Lucent (Nuage Networks, Cloudband).
Nokia also enjoys a full portfolio of networking assets under one roof
(routing, optics, fixed, wireless), which allows it to play a leading role
in NFV and SDN and benefit from a full understanding of the complete
telecom network. We also see Nokia’s credibility in consulting and
system integration as significantly enhanced through the combination.

20 May 2016
Technology
Telco Cloud
Finally, we note that Nokia’s #1 position in Cloud RAN should enable

Nokia to take share with significant enhancements recently launched
via the latest AirScale line. This should be a strong point of
differentiation as C-RAN technology emerges.
Ongoing bull points
 Separately from the cloud discussion, we continue to see Nokia’s
management track record (+15% gross margin from 2011-15 vs
Ericsson at 0%) as a key plus, given the ongoing restructuring
following the acquisition of Alcatel-Lucent. We see Nokia as
determined to bring its Networks margin back from “>7%” this year
and then above 10% next year. We see a 13% Networks margin as
credible by 2018.
 In addition, Nokia is set to approve the resumption of a €1.5bn
buyback at its AGM on June 16th, while a combined €0.26 per share
dividend is due in early July (€0.16 ordinary, €0.10 special), equivalent
to a 5.6% yield.
 Given all of the above and a stronger net cash position, we continue to
see Nokia as meriting a 15% premium over Ericsson, yet the stock
trades on a 7% discount today (EV/EBITDA 2018E: Nokia 4.8x,
Ericsson: 5.1x). Our target of €6.2 offers 33% upside.
 Ericsson (Hold, SKr69 target): Ericsson has 66,000 services
professionals compared with Nokia at likely around 36,000 following
restructuring. Its scale advantage in services is significant, and reflects
the fact that along with Huawei, Ericsson reaches into very small
operators in many countries out of which Nokia has long pulled out.
This should put Ericsson in a strong position to take a leading role at
many tier-2 and tier-3 operators to do deals of the sort they have done
with Swisscom (full stack, full NFV platform). However, we see its
scale in services as a potential risk down the line, given that the
transition to cloud empowers operators to consider the likes of
Accenture and HPE, and then even if they go with Ericsson, they can
more easily pick and choose other vendors to avoid vendor lock-in. We
see the “we’ll do everything for you” model as less viable for tier-1
operators, while overall managed services and network rollout sales
are likely to be vulnerable long-term. Further, on the technology side,
we believe that Ericsson is to date behind others in adoption of C-RAN,
NFV and SDN, while we hear frequent rumblings that Ericsson is less
than enthusiastic about going truly “open”, which is a concern to us.
The muddled Cisco partnership
 Furthermore, Ericsson’s partnership with Cisco appears very
asymmetric in favour of Cisco, which also raises a concern for us
(Cisco gets a channel for its routers/switches, while possible benefits
to Ericsson accrue further out). Ericsson’s main potential benefit from
the partnership would be to cut opex, especially R&D for routing, but
this seems further out than we initially thought with commitments to
customers requiring Ericsson to keep its ‘legacy’ routing portfolio (SSR
8000) alive for longer. A full migration of all functions seems far away
and will likely not happen before 5G arrives. In addition, it remains
unclear how a partnership would work here given Cisco may decide to
push its own EPC from the Starent acquisition in 2009 over Ericsson’s
solution.
 Spirent (Hold, 70p target): Spirent is well positioned for certain aspects
of NFV/SDN testing and offers these capabilities as part of its
TestCenter and Avalanche product lineup. Test solutions include
OpenFlow controller emulation, VM live migration testing, virtual
switch and application testing as well as security tests. However, the

20 May 2016
Technology
Telco Cloud
move from testing individual ‘boxes’ such as routers and switches for
performance towards testing end-to-end software defined networks
and ‘commodity’ off the shelf server hardware is likely going to reduce
demand for purpose-built test tools, in our view. We believe Spirent
needs to continue to migrate its business model further towards
software solutions for NFV/SDN networks and end-to-end service
quality testing while further deemphasizing testing individual kit for
networks in order to remain relevant longer-term, a strategy already
partially pursued by management.
 Ultimately, we believe that Spirent’s future business model may trend
towards a pure licensing business to the larger OEMs given the
disruptions outlined above, and given the increased preponderance for
live testing.
 ADVA (Hold, €10 target): ADVA has a toe-hold in the NFV market with
its Overture acquisition in January 2016. Overture was a well-regarded,
early-stage orchestration vendor (Ensemble), which came combined
with an Ethernet-over-copper business that brought ADVA some key
US customers including Verizon. ADVA’s strategy appears similar to
that of Ciena and its acquisition of Blue Planet. Both companies seem
to be positioning themselves as the “last box standing” with their
Ethernet demarcation device, while other routing and firewall
functions are virtualized. ADVA’s play today in this market has been
the ADVA FSP 150-GE110Pro, which combines carrier Ethernet and IP
services in one element. How this market plays out remains to be seen,
but the key debate remains as to what gets virtualized on premises
and what gets hosted in the network, similar to the broader debate
around virtualized CPE.
 Overall, we believe that ADVA has picked up Overture’s additional
capabilities at a cheap price (US$35m), so we believe that ADVA can
re-position much of this business to support the core business even if
the market does no develop as planned. Our investment concerns on
ADVA relate more to near-term margin trends, driven by a rising
proportion of Web2.0 customers in the mix. However, we concede
that the company is a credible acquisition target.
Implications for the US networking equipment universe

 In our view, the switching and routing vendors - Cisco and Juniper By Vijay Bhagavath & team
(both Buy rated), have a multi-year thematic opportunity in
telco/carrier transformations, driven by the migration to a software-
based and highly automated services delivery infrastructure. A core
software-based network services delivery and automation framework
is at the center of any NFV cloud initiative.
 Cisco and Juniper, in our view, are in the early innings of getting
design wins for their cloud automation and service orchestration
frameworks - e.g. Juniper Contrail, Cisco APIC (Application Policy
Infrastructure Controller), as the core SDN and NFV platform - for
enabling automated and dynamic delivery of network services such as
business VPNs, Ethernet and optical connectivity, managed security
services etc.
 Cisco (Buy, $35 target): Our industry conversations with the
telco/carrier IT channel suggest that Cisco is in the early innings of
selling turnkey telco cloud solutions for web-based services ordering
and automated fulfillment for network service offerings such as layer 3
business VPNs, Ethernet and optical connectivity, managed security

20 May 2016
Technology
Telco Cloud
services, etc. These turnkey engagements are typically multi-year

opportunities for Cisco, involving sales of edge and core routers,
metro optical, software automation and service cataloging solutions,
and software- based mobile packet core, IP video delivery etc.
 Juniper (Buy, $30 target): For Juniper, our view is very similar to Cisco.
We believe Juniper is leading the customer dialogue with their Contrail
NFV cloud software automation solution - so as to pull forward multi-
year sales for their edge routing, switching and network firewall
platforms, and related network integration opportunities. The publicly
announced Contrail design win at AT&T Domain 2.0 for NFV is a case
in point.
 Bigger picture, we see Cisco and Juniper, in particular, increasingly
leading with their telco cloud automation frameworks to create a
"sticky" multi-year order pipeline for their networking platforms at the
major telcos and cable companies over the next few years

20 May 2016
Technology
Telco Cloud
Implications for Telcos

Positive impact on telecom operators from network cloud
Moore’s law and efficiencies in signal coding and greater spectrum availability By Robert Grindle & team
have materially continue to reduce the unit costs of deploying mobile data
capacity. We estimate that operators like Vodafone are seeing unit capacity
costs falling in excess of 30% annually (see below) - just as well given high
levels of consumer demand and the need to offset legacy voice revenues.
Figure 58: Vodafone Group incremental traffic added Figure 59: Vodafone Group unit capex trends
700 40.0
596 34.3
600 35.0 -30%
-28%
Incremental traffic in 12m (PB)
CAGR
500 +77% 30.0 28.3 CAGR
24.8
Capex per GB (£)

CAGR 25.0 -35%
400
+79% 317 334 CAGR 20.0
+75% 279 CAGR 20.0 18.0
300 16.9
CAGR
186 190 15.0 12.8
200 148 11.5
91 99 10.0 7.6
100
5.0
-
Europe AMAP Group -
Europe AMAP Group
12m to Sep 13 12m to Sep 14 12m to Sep 15
12m to Sep 13 12m to Sep 14 12m to Sep 15
Source: Deutsche Bank Source: Deutsche Bank
The DB telecom services research team view the ‘cloudification’ of telco

network resources and the associated disaggregation of hardware and
software supply as a the next stage in an ongoing process of efficiency gains
for the industry, akin to switch digitalization, fiber optics, or in purely economic
terms, the ascendance of Chinese vendors. All telcos should benefit from the
efficiency gains, but larger operators, and those with cross-border footprints,
may prove to be the earliest beneficiaries. Anything that improves the ability of
telcos to respond to changing customer demands is further positive.
Globally there appears to be more discussion about network virtualization and

SDN in the US and Asia, rather than in Europe, perhaps as European operators
are still focused on rolling out 4G coverage rather than leveraging the full-suite
of benefits from future technology releases. Mentions of NFV and SDN are
growing in Europe however with Vodafone recently (FY-16 results) targeting
50% of network functions virtualized by 2020.
NB-IoT revenue opportunities

We would also highlight that Vodafone recently announced a target to have all
4G sites supporting NB-IoT by 2020. This suggests that there will be a
significant software-led upgrade ahead for its LTE network, to drive leading
performance. New vertical-oriented revenue streams continue to have good
potential as revenue growth opportunities for telcos.

20 May 2016
Technology
Telco Cloud
Figure 60: Vodafone’s strategy to maintain leadership
Source: Vodafone
Figure 61: Cloud, SDN, Virtualization as key to network competitiveness
Source: Vodafone
Separately, Deutsche Telekom is upgrading its entire European fixed local loop
to IP, in order to avail of the lower costs and improved agility of
centralized/cloud-based service deployment. Telefonica is also very active in
the telco network cloud standards setting process and like DT, has an all-IP
network strategy, is trialing automized virtual network functions, and has
completed an SDN-IP trial in Peru. In the US, AT&T appears a first mover and
has highlighted Project Agile efficiency initiatives which include “SDN
transformation to drive continued cost momentum”.

20 May 2016
Technology
Telco Cloud
Figure 62: Deutsche Telekom’s network cloud strategy (1)
Source: Deutsche Telekom
Figure 63: Deutsche Telekom’s network cloud strategy (2)
Source: Deutsche Telekom

20 May 2016
Technology
Telco Cloud
Implications for ARM

Key points
 Next-generation networking represents one of the few remaining
growth markets in semiconductors given the maturity of the
smartphone, tablet and PC markets
 The promise of NFV, SDN and C-RAN is leading to significant re-
designs of classic networking equipment to support these concepts
 Almost all semiconductor vendors other than Intel are choosing to
licence ARM’s architecture, as opposed to Power and MIPS in the past
 We view ARM is poised to see its networking unit share rise from 15%
today to 51% in 2020, while Intel should also gain significant share
 Intel could become more of a force in the data plane from 2020
onwards, led by the acquisition of Altera
Networking in the processor architecture landscape
With all the disruption being caused by telco cloud, next-generation

networking (along with autonomous cars, virtual reality) is now one of the few
growth markets left in semiconductors. The once growing markets of PC and
smartphones have matured, leaving the “embedded” market as the key
battleground between ARM and Intel.
Figure 64: Processor landscape, 2014
Market size: $31bn Processor market size: $112bn Market size: $37bn
Other
2% Other
1%
Mobile
phone & PCs &
tablet 27% servers
ARM 34% x86
98% 99%
Intelligent
systems &
ARM victorious, embedded Intel victorious,
Intel on fringes systems ARM on fringes
39%
Other
8%
Market size: $44bn
MIPS ARM
37% The new battleground
20%
of "ARM vs Intel"
Power
x86
19%
16%
Networking is ~30%* of this
Source: Deutsche Bank; IDC’s measure of the wired communications TAM is $9.6bn in 2014 is lower than ARM’s estimated TAM of $13bn in 2015. We believe this relates to ARM including storage

20 May 2016
Technology
Telco Cloud
The chart above shows the processor architecture landscape as it stands as

ARM and Intel face off. In their home turf, each is dominant. However, the
remaining embedded systems area (auto, next-gen networking, industrial IoT
etc) is “up for grabs” and has become the remaining battle ground for share
gain, in the absence of growth. However, this market is much more
fragmented than their home turf with much more customization required. That
has necessitated ARM investing in opex (ARM engineering headcount up to
2,639 in 2015 from 1,803 in 2013), partly to meet the challenge, while 25% of
ARM’s technical engineers are now dedicated to networking. Intel has also not
stood still – it acquired Altera for $17bn, specifically to bulk up in networking.
Instruction set architectures (ISAs) in networking

We show in Figure 65 a chart which gives a high-level summary of the
different CPU and throughput requirements for different networking
applications. The heterogeneity of these applications means that a “one-size
fits all” approach has not historically been the right approach. In fact, there are
a huge number of niche vendors that do well in this market. To date, the MIPS
and Power architectures have been dominant, but Intel and ARM are
challenging hard.
The good news for ARM and Intel is that the move to NFV/SDN and 5G is
driving a faster transition where the partition between the end device, the
network and the cloud is blurring. This is necessitating more intelligence and
analytics to be placed around the network, what ARM calls the “intelligent
flexible cloud”. Much of the functionality in control plane intensive applications
(eg OSS/BSS, layer 4-7 services) are moving into software, based on x86 COTS
(commercial off the shelf) hardware, while the move to 5G and NFV/SDN is
necessitating more intelligence to be placed around the network.
We also show the markets where ASSP/ASIC dominate, given that this
represents the opportunity for ARM to take share at the expense of Power and
MIPS. Key ASIC players include Broadcom/Avago, Cisco, but the boundary
between ASSP and ASIC is becoming increasingly blurred for vendors like
Broadcom and Cavium. Key ASSP players include: Broadcom, Cavium,
Marvell, and Texas Instruments. Multi-core network processor vendors include
NXP, Cavium, Intel, Applied Micro, Broadcom and Mellanox.

20 May 2016
Technology
Telco Cloud
Figure 65: Network processors – instruction set architectures (ISAs)

ASIC/
High ASSP
Power
CPU
OSS/BSS, Wireless
subsystem and GWs* ARM
network control
MIPS
Service
x86 Appliances
Dedicated
appliances (L4-L7) † ASSP
ARM
phased out
CPU
Reqs MIPS ASSP
Wireline Power
GWs** Core backbone
routing ASIC
ASSP Edge Customer edge
x86 routing
access
MIPS
Aggregation and data
Business CPE center switching NPUs
MIPS
L2 switch silicon
Home CPE ARM
Low
0 10Mbps 100Mbps 1Gbps 10Gbps 100Gbps 1Tbps 10Tbps 100Tbps 1Pbps

Throughput
† Services sitting on top of the network, eg load balancing, firewall, * Examples include enterprise LAN gateways, we include base station ICs; ** Wireline gateways include softswitches, media gateways
ARM is likely to take a big share in networking

ARM started targeting the networking space in 2008 and by 2012, ARM v8
licensing (64-bit) began to come through meaningfully into ARM’s licensing
revenue. However, royalties have taken time to come through and as of 2015,
ARM’s share in networking is still only 15%. And perhaps alarmingly, ARM’s
networking units grew only 10% yoy in Q1-16, a marked deceleration versus
the 30% yoy growth posted in Q2-15, and distinctly inferior to 60% growth in
networking at Intel in their Q1-16.
However, we are confident in ARM’s ability to see a re-acceleration in
networking growth (taking share along with Intel), as ARM targets unit share
of 45% by 2020 (DBe: 51/41% unit/value share). The key driver of recovery in
2016 will be ARM v8 licensees (Hisilicon, Broadcom, Cavium, NXP) ramping up
14/16nm finFET processors in 2016, given more attractive yields compared
with 2015. These launches (driven by adoption of NFV/SDN) should help to
re-accelerate unit growth while the average royalty rate should rise
considerably given that many of these chips are not only 64-bit, but many-core
(up to 100). In our view, ARM has done all it can as the design activity is
“baked” while it has gained virtually all of the key ASSP/ASIC vendors in
networking. The market share gains should come at the expense of Power and
MIPS (owned Imagination Technologies), with Power replacement happening
most strongly in the control plane. So between now and 2020, the biggest
variables in determining the trajectory of market share growth are for the most
part outside of ARM’s control and will likely be i) product launch timing, ii) the
rate at which legacy products will be ramped down, iii) adoption of
next-generation networking technologies (including NFV/SDN), and
iv) 5G adoption. What ARM can do to facilitate its growth is in invest in the
software ecosystem and its efforts through Linaro will be a key focus.

20 May 2016
Technology
Telco Cloud
Figure 66: ARM share in networking (2015-2020E)
ARM networking unit share (%)

ARM networking value share (%), DB estimated TAM
ARM networking value share, ARM estimated TAM
60%
51%
50% 44%
40% 37%
29% 41%
30% 35%
22%
29%
20% 15%
23%
10% 17%
11%
0%
2015 2016E 2017E 2018E 2019E 2020E
Figure 67: ARM unit share by networking sub-vertical
Base stations Carrier infrastructure

Layer 2/3 switching Routing
Enterprise access points Other networking
100% Blended average
80%
80%
60% 51%
44%
37% 40%
40%
29%
22% 35%
20% 15%
10%
0%
2015 2016E 2017E 2018E 2019E 2020E
ARM’s share opportunity is greatest in base stations ($1bn processor TAM),

where ARM has signed up 5 of the top 6 OEMs in this market, many of whom
use ASICs and their own DSPs (and a few are CEVA licensees), while key ASSP
vendors like Cavium, NXP and LSI Axxia (owned by Intel) are now all ARM
licensees. We believe ARM can reach 80% share by 2020 (ahead of its 75%
share) as vendors migrate to 5G and C-RAN, while small cells (an important
driver of units) should grow in importance, driven by densification efforts. We
also believe that enterprise access points ($2bn TAM, high in units, low in
value) should continue to be an area of strength as key vendors like Broadcom,
Realtek, Hisilicon, and Mediatek all migrate to ARM (mainly from MIPS).

20 May 2016
Technology
Telco Cloud
Figure 68: ARM’s current progress in networking by OEM
Market OEM selling into Current design wins

Wireless access
Wireless
Mobile Backhaul
Wired access
Wired
Aggregation/Core
High-end enterprise/Data centres
Enterprise Low/mid enterprise equipment
Storage and Security
OEM shipping some ARM based equipment ARM design wins but not yet shipping no ARM design wins
Source: ARM; Grey shading implies this is a key target market for ARM in networking
Of the other areas of networking, the innovation cycle is slower, while in some
cases using ARM is less appropriate (eg in fixed access, core routing,
backhaul), but nevertheless ARM should continue to take share at a steady
rate, given the disruption that is being driven by NFV/SDN. The most notable
vendor in this segment is Broadcom, which has already publicly said that the
“MIPS ecosystem is slowly dying and collapsing” (see the Broadcom analyst
day transcript, December 2014). This is not an encouraging statement for MIPS
(owned by Imagination Technologies) given that Broadcom is its largest royalty
payor for MIPS. Broadcom has already migrated to ARM for its StrataGX
(shipping since Q1-16, quad-core ARM v8) and StrataConnect SoCs, while
Broadcom’s multi-core XLP III processor line (currently on MIPS) looks very
likely to follow to support multiple applications including switch control
processing. However, we caveat that a large portion of Broadcom’s revenue
today is in layer 2/3 switch silicon (Trident/Tomahawk) which does not support
an operating system. Here, ARM’s opportunity is likely to be limited to adding
intelligence adjacent to the switch silicon – ARM claims to be making good
progress here.
Multi-core comms processors

In general multi-core processors ($2bn TAM), ARM has now caught up with
the Power architecture, which had been the workhorse for more control-plane
intensive applications. The best evidence here is that the largest player here,
Freescale (NXP) is now supporting both platforms for its QorIQ processor, but
in our view appears ready to phase Power out, which is remarkable since
Motorola/Freescale was a lead user of the Power architecture (originating out
of IBM). Perhaps more important though will be what NXP does with this
business since it appears to be “for sale” based on the way NXP management
is messaging. Of the other vendors in comms processors, Cavium has
launched a number of ARM processors including ThunderX. In May 2016, it
launched OCTEON TX, which is a complete line of ARM-bsaed SoCs for control
and data plane applications. This represents another loss for MIPS, which had
been in the OCTEON III, found in base stations. Applied Micro is another that a
user of the Power architecture that is migrating to ARM.

20 May 2016
Technology
Telco Cloud
Figure 69: Example ARM-based networking products

Customer (rank in networking) Product Architecture Application
Altera (acquired by Intel) Arria V, Cycle V, ARM v7 (Cortex A9), Wireless infrastructure, aggregation
Arria 10, Stratix 10 ARM v8 (Cortex A53
Applied Micro Helix series ARM v8 Routers, switches
Broadcom (#1) StrataGX, XGS, ARM v7 (Cortex A9), Switch solution, routing, access, SDN switching, Wifi routers, access
BCM4908 ARM v8 points
Cavium OCTEON TX ARMv8 (Cortex A72) Storage, data center, networking (includes control plane application
areas - security, routing, NFV/SDN, CPE, wireless transport, NAS,
storage controllers, IoT gateways, printer and industrial applications
Conexant CX94610 ARM10 Wireless router, gateways
Fujitsu N/A ARM v7 Base stations
Hisilicon (#7) N/A ARM v8 (Cortex A57) Wireless infrastructure, networking, set-top boxes, routers
Intel (#3) Axxia 4500 Cortex A15 Switches, routers, wireless base stations
Marvell (#4) Armada series Cortex A9 Network attached storage devices
MediaTek MT6723 Cortex A7 Gateways, hotspots, IoT devices
NXP (#8) QorIQ series Cortex A7, A53, A53 Routers, network attached storage, wireless access, network edge
Texas Instruments (#2) KeyStone Cortex A15 Small cells, wireless backhaul
Xilinx (#6) Zynq Cortex A9 Wireless communications, data centers, SDN
Competitive environment - a “battle royale” is looming

For now, both ARM and Intel are likely to gain share in networking. ARM in
ASSP/ASICs should be at the expense of Power (more control-plane biased)
and MIPS (mix of control and data plane). Intel’s growth recently has been via
traditional server sales (mostly HP), which telcos are adopting to support new
cloud architectures. ARM however has seen its growth (mainly ARM v7 up
until today) impacted by the general slowdown in the markets it has
succeeded in to date (small cells, enterprise access points, customer premise
equipment (CPE)). Most of this market is suffering because of enterprise and
telecom spending trends, but there are also vendors backing out of the
networking market (eg Texas Instruments and possibly NXP in future), which
makes ARM more sensitive to market trends today.
Intel has made networking a key focus in recent years and the fruits are paying
off (+60% growth in the last quarter), and we think their growth is more limited
by the applicability of the general purpose x86 CPU, rather than the whims of
the market. Intel has acquired aggressively in the market to develop domain
know how (Lantiq, Mindspeed, LSI Axxia ASSP) to expand its SAM. This new
capability is key as today Intel lacks the ability to support data plane intensive
applications that have stringent telco-grade, real-time requirements. For
example, DSPs can outweigh the benefits of using CPUs in infrastructure for
voice or video. However, this is changing and proof points are emerging (eg
cloud RAN), even if a common complaint is that Intel cores are “heavy”.
Intel’s aggressive move to acquire Altera (for $17bn, at a rich price of 26x ex-
cash 2016E EPS) should theoretically put Intel in a position to challenge ARM’s
customer base for these use cases as well. This is where we believe there
should be a “battle royale” in future years between Intel on one side and the
remaining major ARM vendors electing to play in this market (likely Broadcom,
Hisilicon, Cavium, Qualcomm, Xilinx), and potentially OEMs on the hardware
side such as Cisco, Ericsson, Nokia, although that group are less wedded to
silicon development on a 10-year view. Intel’s progress in silicon photonics
could be an important way to cement their strength in networking and the data
center.

20 May 2016
Technology
Telco Cloud
Altera (the #2 FPGA vendor after Xilinx) should put Intel in a position to
integrate a CPU and an FPGA onto a single piece of silicon by 2019 in our
view. That could be a compelling offering to address its inadequacies to
support data plane intensive applications, effectively hardware
programmability and reconfigurability at high speed, a play that could make
ASIC/ASSPs shrink over time. For example, one avenue could be network
processors, which today are used in many of the most intense data plane use
cases as a high-performance and low-power solution for packet handling in
routers, gateways, switches and other network devices. To date, throughput
was the key criteria, but Cisco’s migration of EZchip to an ASIC suggests that
even this is changing. Another key target application is the base station as it is
feature rich and requires re-configurability. We fully expect Intel to be directly
head-on for such business with the Caviums of this world by 2019.
The attractions of FPGAs

FPGAs (field programmable gate arrays) are specialized chips that are
programmed to perform very specific functions in hardware. An FPGA is
effectively a piece of programmable logic. So rather than executing a function
in software, the same function can be implemented in an FPGA and executed
in hardware. This contrasts with ASICs, which are manufactured with custom
designs created by the customer. The main advantage of an FPGA over an
ASIC is that is flexible (time-to-market) and configurable, so that customers
can implement of re-program the logic in an FPGA for a specific feature or
capability that a telco requires. The main disadvantages versus ASICs are
relative cost, higher power consumption and issues around de-bugging. FPGAs
were traditionally used to implement limited functions as “glue logic” to
interconnect on-board devices. However, rising complexity and an increased
desire for programmability has made FPGAs a more attractive target for many
projects in networking.
Figure 70: FPGAs vs ASICs vs ASSPs

ASIC ASSP FPGA
Customisable Yes, by chip fabrication facility No Yes, by end user
Erasibility/re-programmability No No Yes
Relative time to market Slow Immediate Fast
Relative unit cost Low Moderate Moderate to high
Customer’s development cost High Low Moderate
Field upgradability No No Yes
Source: Altera 10-K
For Intel in turn, servers are getting more network-centric, driven by NFV/SDN.
So Intel is both defending its position in servers from ARM’s ambitions (which
we think will disappoint), and targeting growth in networking. What is
tantalizing about silicon integration of CPU plus FPGA is that silicon integration
(for example at 7nm, versus an ASSP/ASIC at 28nm) could solve the latency
challenges of a two-chip package. As things stand, Intel will push a
2.5D/interposer solution at first, but we believe that partitioning the due puts a
lot of constraints on the system. A key ambition for the next three years before
they tape out a single chip solution should be to ramp up the development
community to buy into the FPGA/CPU concept. Traditionally FPGA
development teams have been separated off given their specific programming
skills, so the challenge for Intel is to make these teams drive towards a
common goal.

20 May 2016
Technology
Telco Cloud
Won’t the Altera gains be too late for NFV/SDN?

While Intel’s new solution for networking won’t likely be ready for launch until
2019, Intel still has plenty of opportunity to grow in the first wave of NFV/SDN
via COTS servers, and with its acquired assets which give important customer
relationships at OEMs. Following 2019, we expect Intel’s moves to participate
in the second wave, while their ambition may to hasten the end for
semiconductor design at OEMs, as they look to hope to differentiate in
software instead. That makes their growing relationships with the likes of
Cisco, Ericsson and Nokia very important to their success at beating off the
likes of Cavium. We would not be surprised to see Intel trying to get even
closer to OEMs by acquiring out talent for example in order to secure business.
Finally, it is important to note that Intel completely dwarves many ARM

competitors in terms of R&D. Intel’s product R&D is likely to be $10bn per
annum going forward of which we would argue that at around $2bn is
currently spent on its ambitions in next-generation networking. Only Broadcom
can match this spend at $2bn in total R&D, while small vendors like Cavium
($0.2bn) and Applied Micro ($0.1bn) seem ill-placed to take on Intel long-term,
especially with fast-rising design costs. In this vein, we continue to expect
more consolidation in the networking semiconductor market, which could
restrict the future ARM licensing opportunity. While we also believe that ARM
is having to take on some of the development burden to help its networking
customers get products to market. Some examples of notable deals are below:
Figure 71: Notable M&A in networking semiconductors

Date Acquiror Target Valuation (USDbn)
May-16 Maxlinear Broadcom (backhaul assets) 0.1
Apr-16 Cypress Broadcom (IoT assets) 0.6
Feb-16 Avago Broadcom 37.0
Feb-16 Mellanox EZChip 0.8
Jan-16 Microsemi PMC Sierra 2.5
Dec-15 Intel Altera 16.7
Aug-15 Microchip Micrel 0.8
Aug-15 Qualcomm Ikanos 0.1
Apr-15 Microsemi Vitesse 0.4
Apr-15 Maxlinear Entropic 0.3
Feb-15 Intel Lantiq N/A
Oct-14 Inphi Cortina Systems 0.1
Aug-14 Avago PLX Technology 0.3
Aug-14 Intel Avago/LSI Axxia 0.7
Jul-14 Cavium Xpliant 0.1
Jul-14 EZChip Tilera 0.1
May-14 Avago LSI 6.6
Dec-13 Intel Mindspeed N/A
Jan-12 Intel Qlogic’s Infiniband 0.1
Jul-11 Intel Fulcrum N/A
So in summary, from 2019 onwards then, we expect ARM and Intel to more
directly compete, but for now both look set to succeed. We expect Intel to
grow its share to 25% by 2020, while ARM should grow to 41% (both in value
terms). See Figure 72.

20 May 2016
Technology
Telco Cloud
Figure 72: Market share shifts in networking (by value)
45%
41%
34%
ARM
Intel
25%
19% MIPS
8% 12% Power
8% Others
2014 2016E 2018E 2020E
Networking TAM does not grow by much

We show below our view of the networking silicon TAM, as relevant to ARM
architecture. It is worth pointing out that ARM will likely be significantly
affected by any further slowdown in the networking silicon market. Capex
trends of late have not been supportive in either wireless or wired capex, so
ARM’s progress could be slowed by exposure to a slow-growing market.
Figure 73: Networking silicon TAM, $bn,
20
18
15.9 16.2
15.1 15.3 15.6 Other networking
16 14.8
14
Enterprise access
12 points
10 Routing
8
Layer 2/3 switching
6
4 Carrier
infrastructure
2
Base stations
0
2015 2016E 2017E 2018E 2019E 2020E
We also note that there is some debate about the relative size of the TAM with
IDC suggesting the wired processor TAM was $10bn in 2014, while Gartner
suggests it is $11-12bn. ARM appears to be at $13bn for 2015, but this
estimate appears to include a contribution from the storage market within its
definition of “networking”.

20 May 2016
Technology
Telco Cloud
ARM in enterprise networking

We should note that ARM has a significant opportunity in enterprise
networking, which we have not really addressed yet. The key point here is that
the cloud service providers (Amazon, Facebook, Google, Microsoft et al) are re-
defining how future data centers look. Their capex as can be seen in Figure 74
is enormous and growing fast. They are not just requesting customized CPUs
from Intel, but they are also customizing the two-thirds of hardware spend that
is not servers, for examples, in areas like switch fabric, security appliances,
load balancers, and firewalls. They also are extending into optics. So here,
ARM potentially has an opportunity accelerate the migration of ASSP/ASIC
silicon that currently uses Power and/or MIPS, often ultimately selling into the
likes of Cisco. The key customers would be Cavium, NXP et al. While Intel too
is hoping to disrupt these segments, we would expect Intel’s competitors to
continue to migrate over to ARM (whilst potentially losing share to Intel along
the way). What will decide their success or failure will be the actions of the
large Web 2.0 vendors which are ultimately creating proprietary silicon in
many cases. ARM is often the only choice outside of server silicon for these
buyers as they go more vertical in the data center. This is particularly true in
SDN-type applications. Given that we believe that Amazon, Microsoft, Google
et al are all ARM licensees, ARM is in a strong position to play a role here,
even if it struggles to take significant share against high-end Xeons from Intel.
Figure 74: Cloud service providers (the big 8) - capex ($bn)
Apple Google Amazon Microsoft

Facebook Tencent Alibaba Baidu
48
45
37 38
27
22
15
11
2010 2011 2012 2013 2014 2015 2016E 2017E
Source: Company data, Thomson Reuters, Deutsche Bank estimates
A note on ARM in servers – a David vs Goliath story

We believe that ARM will only occupy 11% of its stated server TAM of $20bn
by 2020 (ARM’s unit target is 25% while IDC estimates they can achieve 7%),
given that the ASP of its customers’ silicon is likely to be much lower than an
Intel Xeon E7. We also believe the 25% unit share, if achieved by 2020, would
not translate into more than 10-15% value share of this market. While design
activity is high at key customers like Cavium (ThunderX), AMD (Seattle),
nVIDIA (Denver), Broadcom (Vulcan), and at Qualcomm, ARM’s success in
servers also heavily relies on China through Hisilicon (Huawei) and Phytium.
We continue to believe that ARM’s opportunity is fairly limited to the low-end
given insufficient performance and the high demands of most server
workloads. So for now, we believe that the latest wave of ARM server
products will not take significant share, despite the encouragement of China
and OEMs like HP and Dell (who wish to squeeze Intel on price).

20 May 2016
Technology
Telco Cloud
Feedback to date we hear from server OEMs on first ARM-based chip launches
such as Cavium’s ThunderX has been mixed, with performance decent, but
power consumption not yet being meaningfully better than what Intel has to
offer. Server OEMs seem to remain in a ‘wait and see’ mode for now with
more ARM-based silicon scheduled to come to market in 2017. Longer-term,
ARM’s best hope is probably via the cloud service providers (Tencent, Alibaba,
Amazon, Google et al). Facebook remains very influential given their role in the
Open Compute Project. Cavium and Applied Micro meanwhile remain the most
invested companies to date in ARM servers and yet their combined annual
R&D is only $0.3bn to support all of their server and networking efforts. This is
likely 20x smaller than Intel. Cavium meanwhile is only targeting 50-100k
shipped units by H2-17, less than 1% of the server market.
Figure 75: ARM server products (2015-18E)

Date Company Product Node
H1-15 Applied Micro X-Gene 2 28nm
H1-15 Cavium ThunderX 28nm
H2-15 AMD Seattle (A1100) 28nm
H1-16 Applied Micro X-Gene 3 16nm
H2-16 Cavium Thunder X-2 14nm
H2-17 AMD AMD K12 14nm
2018 Broadcom Vulcan 16nm
2018 Qualcomm TBD 14nm
Abandoned – 2014 nvIDIA Denver
Closed -2013 Calxeda
We increase our target price on ARM to £10.50
While we are positive on ARM’s prospects in networking given clear share

gains, the fact remains that ARM’s networking business will account only 12%
of PD royalties by 2020. So networking is too small to move the needle on
near-term estimates. However, it remains an important driver of growth in non-
Mobile (currently 30% of PD royalties), and the DCF valuation. The networking
and server opportunities have already driven a considerable increase in opex
(up 55% from 2014 to 2016).
We note that Mobility (smartphone and tablets) accounts for 60% of royalties
even in 2020 (versus 70% today), so ARM’s position in smartphones will
remain the #1 driver of the share price. We currently forecast Mobility-derived
royalties to grow at 7% CAGR in USD from 2016-20E while non-Mobile-derived
royalties are set to grow by 20% CAGR in USD.

20 May 2016
Technology
Telco Cloud
Figure 76: ARM: Networking to represent 12% of PD royalties by 2020
1,400
1,200 Others
1,000
Embedded
800
Servers
600
Networking
400
200 Smartphones,
tablets and
0 notebooks
2015 2016E 2017E 2018E 2019E 2020E
So given a more optimistic view on ARM’s share gains in networking, we lift

our price target on ARM, given that we have lifted our long-term royalty
assumptions on ARM’s networking business. Our DCF-derived valuation has
risen to £10.50 (from £10). Our DCF includes the following assumptions:
WACC: 8.5% and terminal growth of 4%.
Our rating remains Hold on ARM

We keep a Hold rating because of our concerns on mobile. With the global
smartphone market approaching zero growth and ARM’s structural royalty
drivers in mobile slowing, we see a sustained slowdown of the company’s
royalty growth ahead. While we are positive on ARM’s outlook in networking
for the coming years, solid progress here will not be enough in our view to
return to above mid-teens USD processor royalty growth over the coming
years. This compares to a close to 20% CAGR from 2012 to 2015. ARM should
continue to see decent earnings growth at a 13% 2015-18E CAGR but we see
a risk that this gets largely offset as ARM continues to ‘grow into’ its multiple
of still 27x 2017E P/E, meaningfully above peers.
We downgrade Imagination Technologies to Sell
We have been critical of Imagination’s strategy over the last years, arguing
that under-investment vs ARM and spreading R&D too thinly across too many
different areas will not lead to market share gains. This thesis has played out
so far with ARM keeping design wins at Samsung LSI, MediaTek and other
previous Imagination customers in mobile graphics. The company is now left
with high exposure to its key mobile graphics customer Apple (DBe ~50% of
royalties) and the acquired MIPS business (DBe ~25%). Given the challenges
we outlined in this report for the MIPS business and ongoing weak Apple
momentum, we downgrade the shares from Hold to Sell.
Too late for a MIPS turnaround and Apple momentum remaining weak
While we welcome management change at Imagination and view the new
CEO/CFO team in charge as an improvement, we note that semiconductor IP is
a long lead time business and it could be too late for new management to still

20 May 2016
Technology
Telco Cloud
turn around the company’s fortunes. Imagination has invested only half of
ARM’s annual R&D budget over the last years while aiming to compete with
them across mobile graphics, apps processors, networking, set-top boxes,
microcontrollers and many other areas. This level of under-investment
combined with ARM’s efforts to move into the networking space has led key
licensees such as Broadcom and Cavium to de-emphasize the MIPS
architecture. MIPS overall accounts for 25% of Imagination’s royalties and we
believe Networking is a large component of this, while Broadcom is the largest
royalty payor, a company that describes MIPS as “dying”. These royalties are
at risk, in our view, as key licensees have already made a decision to migrate
to ARM, partially for now and likely full over time. Taking into account limited
growth potential with key royalty contributor Apple and recent negative data
points from the iPhone supply chain, we believe Imagination will continue to
suffer from disappointing royalty trends over the next years. Restructuring and
cost cuts likely mean Imagination can avoid heavy losses for some time but we
see the outlook for the business as challenged.
A sale is the last exit? But at what price?

As a fundamental turnaround of Imagination appears highly challenging to us,
a sale of the business could be a way out. In this context, we note that Apple
currently owns ~8% and Chinese holding company Tsinghua recently bought
3% of outstanding shares. However, Apple has also publicly commented on
23rd March 2016 that it had talks with Imagination, but does not plan to make
an offer for the company at this time. Tsinghua remains a potential buyer as
China is looking to build a semiconductor ecosystem for the country. Having
said that, we believe Apple is unlikely to remain an Imagination customer if a
Chinese company acquires Imagination as it would fear loss of intellectual
property. IP providers have early insights into development roadmaps of their
licensees and Apple is unlikely to provide such a level of early insight to
Tsinghua, in our view. We therefore believe that a sale of Imagination to
Tsinghua would only happen if Apple moves away from Imagination. This
would leave the company with heavy losses and exposure mostly to the
challenged MIPS business, something that will likely weigh on a potential deal
multiple if the business gets acquired. Meanwhile Chinese fabless companies
continue to adopt ARM broadly.
We value Imagination’s Technologies business (ex the discontinued PURE

radio division) at 3.25x CY2017E EV/sales while we expect PURE to be sold at
not more than 0.5x. The 3.25x is a 75/25% mix of the ~4x EV/sales multiple pre
FY-16 and the 1x IMG paid for MIPS in 2012. We note that 3.25x for
Technologies already implies a return to >20% operating margin versus DBe of
7.4% in FY-2017E and 12% in FY-2018E. While we cannot judge what strategic
price a potential buyer of the business may be willing to pay, the fundamental
valuation upside for Imagination’s shares seems limited to us. We also note
that at the 1x EV/sales multiple for which Imagination bought MIPS,
Imagination’s shares would only be worth 32p currently (based on 2017E and
assuming a PURE sale).
We remind investors that Imagination Technologies is due to report its trading

update on May 26th and its full year results on July 5th. In the year to April
2016, we expect full year sales of GBP152.4m, an adjusted EBIT loss of
GBP(4.1m), and an adjusted net loss of GBP (1.7m).

20 May 2016
Technology
Telco Cloud
Reference architectures
In this appendix, we provide more information for those who would like to
understand more about key telco cloud concepts, NFV and SDN. Many of the
definitions come from the organizations themselves.
NFV and SDN reference architectures

We show below the key reference architectures below from the key relevant
bodies, namely ETSI for NFV, and ONF for SDN.
Figure 77: NFV reference architecture Figure 78: SDN framework
Operation/Business Support Systems Orchestrator
VNF VNF VNF VNF VNF Manager
Virtual Virtual Virtual

Computing Networking Storage
Virtualized
Virtualization layer Infrastructure
Manager
Compute Network Storage
Shared Hardware Resources
Virtualized Infrastructure
Source: ETSI Source: Open Networking Foundation
The NFV architecture is comprised of:
 NFVi (NFV infrastructure): NFVi is the totality of all hardware and

software components which build up the environment in which virtual
network functions (VNFs) are deployed. The NFV infrastructure can
span across several locations.
 VNFs (Virtualized network functions): A VNF is a virtualization of a
network function in a legacy non-virtualized network, such as MME
(mobility management entity) in 4G network.
 NFV-MANO: NFV-MANO refers to the collection of all functional
blocks, data repositories used by these functional blocks, and
reference points and interfaces through which these functional blocks
exchange information for the purpose of managing and orchestrating
NFVi and VNFs.
 VIM: Virtualized infrastructure manager (VIM) comprises the
functionalities that are used to control and manage the interaction of a
VNF with computing, storage and network resources under its
authority.
 Orchestrator: The orchestrator is in charge of the orchestration and
management of NFV infrastructure and software resources, and
realizing network services on NFVi.
 VNF manager: The VNF manager is responsible for VNF lifecycle
management (instantiation, update, query, scaling, and termination).

20 May 2016
Technology
Telco Cloud
The SDN framework is comprised of:

 Application layer: consists of the end-user business applications that
consume the SDN communication services. The boundary between
the Application Layer and the Control Layer is traversed by the
northbound API.
 Control layer: provides the centralized control functionality that
supervised the network forwarding behaviors through an open
interface.
 Infrastructure layer: consists of the network elements and devices that
provide packet switching and forwarding.
An SDN architecture is characterized by three key attributes:
 Logically centralized intelligence: network control is distributed from
forwarding using a southbound interface, OpenFlow. By centralizing
network intelligence, decision-making is facilitated base on a global
view of the network, as opposed to today’s networks, which are built
on an autonomous system view where nodes are unaware of the
overall state of the network.
 Programmability. SDN networks are inherently controlled by software
functionality, which may be provided by vendors or the network
operators themselves. Such programmability enables network
configuration to be automated, influenced by rapid adoption of the
cloud. By providing open APIs for applications to interact with the
network, SDN networks can achieve significant innovation and
differentiation.
 Abstraction: In an SDN network, the business applications that
consume SDN services are abstracted from the underlying network
technologies. Network devices are also abstracted from the SDN
Control Layer to ensure portability and future-proofing of investments
in network services, the network software resident in the Control
Layer.

20 May 2016
Technology
Telco Cloud
Glossary
Key terms
Carrier grade: a system that is very reliable – exceeding “five nines” availability
Control (or signaling) plane: The part of the network that carries signaling
traffic
Cord cutting: the process of cutting cable connections in favour of OTT
Core router: A router that forwards packets to computer hosts in a network
Data (or forwarding) plane: The part of the network that carries user traffic
Diameter signaling: a protocol used for controlling signaling in IMS and LTE
eNodeB: performs radio resource functions such as allocating radio resources
Edge router: A router that routes data packets between a network you control
to a network you don’t
Femtocell: a low-power cellular base station for home or small businesses
Firewall: a network security device that grants or rejects network access
Gateway: a network node equipped for interfacing with another network that
uses different protocols
Hadoop: a free, Java-based programming framework
Layer 1: the physical layer in the 7-layer OSI model
Layer 2: the data link layer in the 7-layer OSI model
Layer 3: the network layer in the 7-layer OSI model
Layer 4: the transport layer in the 7-layer OSI model
Layer 5: the session layer in the 7-layer OSI model
Layer 6: the presentation layer in the 7-layer OSI model
Layer 7: the application layer in the 7-layer OSI model
Layers 4-7: the network services within the 7-layer OSI model, the “upper
layer”
Network slicing: An end-to-end logically isolated network, which includes core,
transport and access functions
OpenStack: an open source, cloud management system
Open source: software whose source code is available for modification or
enhancement by anyone
Picocell: a small cellular base station covering a small area, eg in-building
Power architecture: a RISC architecture that originated from IBM
Router: a device that forwards data packets along network
Server: a computer program or device that provides functionality for other
programs or devices
Small cells: low-powered radio access nodes with a range up to 1-2kms
Switch: a device that channels data from multiple ports to an output port
Virtualization: the act of creating a virtual machine that behaves like a real
computer with an operating system

20 May 2016
Technology
Telco Cloud
Acronyms
A-I
5G: Fifth-generation mobile network

AAA: authentication, authorization and accounting
ANDSF: access network discovery and selection function
AP: access point
API: application program interface
ARM: Advanced RISC Machines
ARPU: average revenue per user
ASIC: application-specific integrated circuit
ASSP: application-specific standard product
BBU: baseband unit
BGP: border gateway protocol
BNG: broadband network gateway
BRAS: broadband remote access server
BSC: base station controller
BSS: business support system (eg billing, customer care)
BTS: base transceiver station
CDN: content delivery network
CE router: customer edge router
COTS: commercial off the shelf, normally relates to standard servers
CPE: customer premises equipment
CPRI: common public radio interface
CPU: central processing unit
C-RAN: cloud radio access network
DNS: domain name server
DPI: deep packet inspection
DLSAM: digital subscriber line access multiplexer
DWDM: dense wavelength division multiplexing
EANTC: European Advanced Networking Test Center
eNodeB: the base station that controls a cell in LTE
ePDG: evolved packet data gateway
EPC: evolved packet core
ETSI: European Telecommunications Standards Institute
ForCES: forwarding and control element separation
FPGA: field programmable gate array
GGSN: GRPS core network
Gi-LAN: gateway internet LAN interface
GPON: gigabit passive optical network
HetNet: heterogeneous network
HSS: home subscriber server
IaaS: infrastructure as a service
IMS: IP multimedia subsystem

20 May 2016
Technology
Telco Cloud
I-S
IP: internet protocol

JSON: JavaScript Object Notation
KVM: Kernal-based virtual machine for Linux
LTE: Long-term Evolution, commonly marketed as 4G LTE
LTE Advanced (LTE-A): an enhancement of LTE, standardized in 3GPP rel.10
LTE Advanced Pro: an enhancement of LTE-A, standardized in 3GPP rel.13
M2M: machine-to-machine
MANO: management and orchestration
MIMO: multiple-input, multiple-output
MIPS: a reduced instruction set (RISC) architecture (owned by Imagination)
MME: mobility management entity
MPLS: multi-protocol label switching
MVNO: mobile virtual network operator
NAT: network address translation
NB-IoT: narrow-band internet-of-things
NFV: network functions virtualization
NFVi: NFV infrastructure
NIC: network interface controller
NID: network interface device
OFCS: offline charging system
OFDM: orthogonal frequency division multiplexing
ONF: Open Networking Foundation
OSS: operational support system (eg fault management)
OTT: over-the-top content (eg Netflix)
PaaS: platform as a service
PCEF: policy and charging enforcement function
PRCF: policy and charging rules function
PCSCF: proxy call session control function
PE router: provider edge router
P-GW: packet data network gateway
QoE: quality of experience
QoS: quality of service
RAN: radio access network
REST: representational state transfer
RISC: reduced instruction set computing
RNC: radio network controller
RPC: remote procedure call
RRH: remote radio head
RRM: radio resource management
RSA: Rack Scale Architecture (Intel developed)
SaaS: software as a service
SBC: session border controller
SDK: software development kit

20 May 2016
Technology
Telco Cloud
S-W
SDN: software defined networking

SD-WAN: software-defined WAN (wide area network)
SFC: service function chaining
SGSN: serving GPRS support node
S-GW: serving gateway
SLA: service-level agreement
SON: self-organizing network
TCO: total cost of ownership
TCP: transmission control protocol
TDM: time-division multiplexing
TE: traffic engineering
UE: user equipment
vBNG: virtualized broadband network gateway
vCDN: virtualized content delivery network
vCPE: virtualized customer premise equipment
vEPC: virtualized evolved packet core
vFW: virtual firewall
vIMS: virtualized IMS
vNAT: virtual network address translation
VM: virtual machine
VNF: virtualized network function
vNPaaS: virtual network platform as a service
VoLTE: voice over LTE
vRAN: virtualized radio access network
WAN: wide area network
WAN acceleration: techniques to up the efficiency of data across the WAN
WDM: wavelength division multiplexing

20 May 2016
Technology
Telco Cloud
Appendix
Important Disclosures
Additional information available upon request
Disclosure checklist
Company Ticker Recent price* Disclosure
ARM Holdings ARM.L 925.00 (GBp) 19 May 16 2,6,9
Ericsson ERICb.ST 62.45 (SEK) 19 May 16 2,14,15
Nokia NOKIA.HE 4.61 (EUR) 19 May 16 2,6,7,9,14
*Prices are current as of the end of the previous trading session unless otherwise indicated and are sourced from local exchanges via Reuters, Bloomberg and other vendors . Other
information is sourced from Deutsche Bank, subject companies, and other sources. For disclosures pertaining to recommendations or estimates made on securities other than the
primary subject of this research, please see the most recently published company report or visit our global disclosure look-up page on our website at
http://gm.db.com/ger/disclosure/DisclosureDirectory.eqsr.
Important Disclosures Required by U.S. Regulators

Disclosures marked with an asterisk may also be required by at least one jurisdiction in addition to the United States.
See Important Disclosures Required by Non-US Regulators and Explanatory Notes.
2. Deutsche Bank and/or its affiliate(s) makes a market in securities issued by this company.
6. Deutsche Bank and/or its affiliate(s) owns one percent or more of any class of common equity securities of this
company calculated under computational methods required by US law.
7. Deutsche Bank and/or its affiliate(s) has received compensation from this company for the provision of investment
banking or financial advisory services within the past year.
14. Deutsche Bank and/or its affiliate(s) has received non-investment banking related compensation from this company
within the past year.
15. This company has been a client of Deutsche Bank Securities Inc. within the past year, during which time it received
non-investment banking securities-related services.
Important Disclosures Required by Non-U.S. Regulators

Please also refer to disclosures in the Important Disclosures Required by US Regulators and the Explanatory Notes.
2. Deutsche Bank and/or its affiliate(s) makes a market in securities issued by this company.
company calculated under computational methods required by US law.
7. Deutsche Bank and/or its affiliate(s) has received compensation from this company for the provision of investment
banking or financial advisory services within the past year.
company calculated under computational methods required by India law.
For disclosures pertaining to recommendations or estimates made on securities other than the primary subject of this
research, please see the most recently published company report or visit our global disclosure look-up page on our
website at http://gm.db.com/ger/disclosure/DisclosureDirectory.eqsr
Analyst Certification
The views expressed in this report accurately reflect the personal views of the undersigned lead analyst about the
subject issuers and the securities of those issuers. In addition, the undersigned lead analyst has not and will not receive
any compensation for providing a specific recommendation or view in this report. Robert Sanders/Johannes Schaller

20 May 2016
Technology
Telco Cloud
Historical recommendations and target price: ARM Holdings (ARM.L)

(as of 5/19/2016)
1,400.00 Previous Recommendations

9 Strong Buy
1,200.00 8 Buy
4 10 12
3 Market Perform
11 Underperform
1,000.00 1
2 5 6 Not Rated
7 Suspended Rating
Security Price
800.00
Current Recommendations
Buy
600.00
Hold
Sell
400.00 Not Rated
Suspended Rating
*New Recommendation Structure
200.00 as of September 9,2002
0.00
May 13 Aug 13 Nov 13 Feb 14 May 14 Aug 14 Nov 14 Feb 15 May 15 Aug 15 Nov 15 Feb 16
Date
1. 25/07/2013: Sell, Target Price Change GBP470.00 7. 22/10/2014: Hold, Target Price Change GBP900.00
2. 30/08/2013: Upgrade to Buy, Target Price Change GBP1,080.00 8. 12/02/2015: Hold, Target Price Change GBP1,080.00
3. 23/10/2013: Buy, Target Price Change GBP1,130.00 9. 22/04/2015: Hold, Target Price Change GBP1,130.00
4. 09/01/2014: Downgrade to Hold, Target Price Change GBP1,070.00 10. 22/07/2015: Hold, Target Price Change GBP1,070.00
5. 05/02/2014: Hold, Target Price Change GBP975.00 11. 12/10/2015: Hold, Target Price Change GBP980.00
6. 23/07/2014: Hold, Target Price Change GBP950.00 12. 27/10/2015: Hold, Target Price Change GBP1,000.00
Historical recommendations and target price: Ericsson (ERICb.ST)

(as of 5/19/2016)
140.00 Previous Recommendations
Strong Buy
120.00 Buy
Market Perform
10
11
Underperform
100.00
8 9
Not Rated
2 3 45 6 7 12 13 Suspended Rating
Security Price
1
80.00
14 Current Recommendations
Buy
60.00
Hold
Sell
40.00 Not Rated
Suspended Rating
0.00
Date
1. 22/07/2013: Buy, Target Price Change SEK85.00 8. 21/07/2014: Buy, Target Price Change SEK98.00
6. 24/04/2014: Buy, Target Price Change SEK87.00 13. 15/01/2016: Downgrade to Hold, Target Price Change SEK85.00
7. 15/07/2014: Buy, Target Price Change SEK85.00 14. 22/04/2016: Hold, Target Price Change SEK69.00

20 May 2016
Technology
Telco Cloud
Historical recommendations and target price: Nokia (NOKIA.HE)

(as of 5/19/2016)
9.00 Previous Recommendations
8.00 Strong Buy

Buy
9 13
7.00 8
7 14 Market Perform
11
4 6 10 Underperform
12 Not Rated
6.00 3
5 15 Suspended Rating
Security Price
5.00 16
Current Recommendations
2
4.00 Buy
1 Hold
3.00
Sell
Not Rated
Suspended Rating
2.00
0.00
Date
1. 02/07/2013: Sell, Target Price Change EUR2.20 9. 30/01/2015: Hold, Target Price Change EUR6.50
2. 03/09/2013: Upgrade to Hold, Target Price Change EUR4.50 10. 01/05/2015: Hold, Target Price Change EUR6.00
3. 30/10/2013: Hold, Target Price Change EUR4.75 11. 03/08/2015: Hold, Target Price Change EUR6.20
4. 09/01/2014: Hold, Target Price Change EUR5.25 12. 21/09/2015: Upgrade to Buy, Target Price Change EUR7.50
5. 30/04/2014: Hold, Target Price Change EUR5.20 13. 30/10/2015: Buy, Target Price Change EUR8.00
Equity rating key Equity rating dispersion and banking relationships

Buy: Based on a current 12- month view of total 350
share-holder return (TSR = percentage change in 54 %
300
share price from current price to projected target price 250 40 %
plus pro-jected dividend yield ) , we recommend that 200
investors buy the stock. 150 51 % 39 %
Sell: Based on a current 12-month view of total share- 100
6 % 31 %
50
holder return, we recommend that investors sell the
0
stock
Buy Hold Sell
Hold: We take a neutral view on the stock 12-months
out and, based on this time horizon, do not Companies Covered Cos. w/ Banking Relationship
recommend either a Buy or Sell.
European Universe
Newly issued research recommendations and target
prices supersede previously published research.

20 May 2016
Technology
Telco Cloud
Regulatory Disclosures
1.Important Additional Conflict Disclosures
Aside from within this report, important conflict disclosures can also be found at https://gm.db.com/equities under the
"Disclosures Lookup" and "Legal" tabs. Investors are strongly encouraged to review this information before investing.
2.Short-Term Trade Ideas

Deutsche Bank equity research analysts sometimes have shorter-term trade ideas (known as SOLAR ideas) that are
consistent or inconsistent with Deutsche Bank's existing longer term ratings. These trade ideas can be found at the
SOLAR link at http://gm.db.com.

20 May 2016
Technology
Telco Cloud
Additional Information
The information and opinions in this report were prepared by Deutsche Bank AG or one of its affiliates (collectively
"Deutsche Bank"). Though the information herein is believed to be reliable and has been obtained from public sources
believed to be reliable, Deutsche Bank makes no representation as to its accuracy or completeness.
If you use the services of Deutsche Bank in connection with a purchase or sale of a security that is discussed in this
report, or is included or discussed in another communication (oral or written) from a Deutsche Bank analyst, Deutsche
Bank may act as principal for its own account or as agent for another person.
Deutsche Bank may consider this report in deciding to trade as principal. It may also engage in transactions, for its own
account or with customers, in a manner inconsistent with the views taken in this research report. Others within
Deutsche Bank, including strategists, sales staff and other analysts, may take views that are inconsistent with those
taken in this research report. Deutsche Bank issues a variety of research products, including fundamental analysis,
equity-linked analysis, quantitative analysis and trade ideas. Recommendations contained in one type of communication
may differ from recommendations contained in others, whether as a result of differing time horizons, methodologies or
otherwise. Deutsche Bank and/or its affiliates may also be holding debt securities of the issuers it writes on.
Analysts are paid in part based on the profitability of Deutsche Bank AG and its affiliates, which includes investment
banking revenues.
Opinions, estimates and projections constitute the current judgment of the author as of the date of this report. They do
not necessarily reflect the opinions of Deutsche Bank and are subject to change without notice. Deutsche Bank has no
obligation to update, modify or amend this report or to otherwise notify a recipient thereof if any opinion, forecast or
estimate contained herein changes or subsequently becomes inaccurate. This report is provided for informational
purposes only. It is not an offer or a solicitation of an offer to buy or sell any financial instruments or to participate in any
particular trading strategy. Target prices are inherently imprecise and a product of the analyst’s judgment. The financial
instruments discussed in this report may not be suitable for all investors and investors must make their own informed
investment decisions. Prices and availability of financial instruments are subject to change without notice and
investment transactions can lead to losses as a result of price fluctuations and other factors. If a financial instrument is
denominated in a currency other than an investor's currency, a change in exchange rates may adversely affect the
investment. Past performance is not necessarily indicative of future results. Unless otherwise indicated, prices are
current as of the end of the previous trading session, and are sourced from local exchanges via Reuters, Bloomberg and
other vendors. Data is sourced from Deutsche Bank, subject companies, and in some cases, other parties.
Macroeconomic fluctuations often account for most of the risks associated with exposures to instruments that promise
to pay fixed or variable interest rates. For an investor who is long fixed rate instruments (thus receiving these cash
flows), increases in interest rates naturally lift the discount factors applied to the expected cash flows and thus cause a
loss. The longer the maturity of a certain cash flow and the higher the move in the discount factor, the higher will be the
loss. Upside surprises in inflation, fiscal funding needs, and FX depreciation rates are among the most common adverse
macroeconomic shocks to receivers. But counterparty exposure, issuer creditworthiness, client segmentation, regulation
(including changes in assets holding limits for different types of investors), changes in tax policies, currency
convertibility (which may constrain currency conversion, repatriation of profits and/or the liquidation of positions), and
settlement issues related to local clearing houses are also important risk factors to be considered. The sensitivity of fixed
income instruments to macroeconomic shocks may be mitigated by indexing the contracted cash flows to inflation, to
FX depreciation, or to specified interest rates – these are common in emerging markets. It is important to note that the
index fixings may -- by construction -- lag or mis-measure the actual move in the underlying variables they are intended
to track. The choice of the proper fixing (or metric) is particularly important in swaps markets, where floating coupon
rates (i.e., coupons indexed to a typically short-dated interest rate reference index) are exchanged for fixed coupons. It is
also important to acknowledge that funding in a currency that differs from the currency in which coupons are
denominated carries FX risk. Naturally, options on swaps (swaptions) also bear the risks typical to options in addition to
the risks related to rates movements.

20 May 2016
Technology
Telco Cloud
Derivative transactions involve numerous risks including, among others, market, counterparty default and illiquidity risk.
The appropriateness or otherwise of these products for use by investors is dependent on the investors' own
circumstances including their tax position, their regulatory environment and the nature of their other assets and
liabilities, and as such, investors should take expert legal and financial advice before entering into any transaction similar
to or inspired by the contents of this publication. The risk of loss in futures trading and options, foreign or domestic, can
be substantial. As a result of the high degree of leverage obtainable in futures and options trading, losses may be
incurred that are greater than the amount of funds initially deposited. Trading in options involves risk and is not suitable
for all investors. Prior to buying or selling an option investors must review the "Characteristics and Risks of Standardized
Options”, at http://www.optionsclearing.com/about/publications/character-risks.jsp. If you are unable to access the
website please contact your Deutsche Bank representative for a copy of this important document.
Participants in foreign exchange transactions may incur risks arising from several factors, including the following: ( i)
exchange rates can be volatile and are subject to large fluctuations; ( ii) the value of currencies may be affected by
numerous market factors, including world and national economic, political and regulatory events, events in equity and
debt markets and changes in interest rates; and (iii) currencies may be subject to devaluation or government imposed
exchange controls which could affect the value of the currency. Investors in securities such as ADRs, whose values are
affected by the currency of an underlying security, effectively assume currency risk.
Unless governing law provides otherwise, all transactions should be executed through the Deutsche Bank entity in the
investor's home jurisdiction.
United States: Approved and/or distributed by Deutsche Bank Securities Incorporated, a member of FINRA, NFA and
SIPC. Analysts employed by non-US affiliates may not be associated persons of Deutsche Bank Securities Incorporated
and therefore not subject to FINRA regulations concerning communications with subject companies, public appearances
and securities held by analysts.
Germany: Approved and/or distributed by Deutsche Bank AG, a joint stock corporation with limited liability incorporated
in the Federal Republic of Germany with its principal office in Frankfurt am Main. Deutsche Bank AG is authorized under
German Banking Law and is subject to supervision by the European Central Bank and by BaFin, Germany’s Federal
Financial Supervisory Authority.
United Kingdom: Approved and/or distributed by Deutsche Bank AG acting through its London Branch at Winchester
House, 1 Great Winchester Street, London EC2N 2DB. Deutsche Bank AG in the United Kingdom is authorised by the
Prudential Regulation Authority and is subject to limited regulation by the Prudential Regulation Authority and Financial
Conduct Authority. Details about the extent of our authorisation and regulation are available on request.
Hong Kong: Distributed by Deutsche Bank AG, Hong Kong Branch.
India: Prepared by Deutsche Equities India Pvt Ltd, which is registered by the Securities and Exchange Board of India
(SEBI) as a stock broker. Research Analyst SEBI Registration Number is INH000001741. DEIPL may have received
administrative warnings from the SEBI for breaches of Indian regulations.
Japan: Approved and/or distributed by Deutsche Securities Inc.(DSI). Registration number - Registered as a financial
instruments dealer by the Head of the Kanto Local Finance Bureau (Kinsho) No. 117. Member of associations: JSDA,
Type II Financial Instruments Firms Association and The Financial Futures Association of Japan. Commissions and risks
involved in stock transactions - for stock transactions, we charge stock commissions and consumption tax by
multiplying the transaction amount by the commission rate agreed with each customer. Stock transactions can lead to
losses as a result of share price fluctuations and other factors. Transactions in foreign stocks can lead to additional
losses stemming from foreign exchange fluctuations. We may also charge commissions and fees for certain categories
of investment advice, products and services. Recommended investment strategies, products and services carry the risk
of losses to principal and other losses as a result of changes in market and/or economic trends, and/or fluctuations in
market value. Before deciding on the purchase of financial products and/or services, customers should carefully read the
relevant disclosures, prospectuses and other documentation. "Moody's", "Standard & Poor's", and "Fitch" mentioned in
this report are not registered credit rating agencies in Japan unless Japan or "Nippon" is specifically designated in the
name of the entity. Reports on Japanese listed companies not written by analysts of DSI are written by Deutsche Bank

20 May 2016
Technology
Telco Cloud
Group's analysts with the coverage companies specified by DSI. Some of the foreign securities stated on this report are
not disclosed according to the Financial Instruments and Exchange Law of Japan.
Korea: Distributed by Deutsche Securities Korea Co.
South Africa: Deutsche Bank AG Johannesburg is incorporated in the Federal Republic of Germany (Branch Register
Number in South Africa: 1998/003298/10).
Singapore: by Deutsche Bank AG, Singapore Branch or Deutsche Securities Asia Limited, Singapore Branch (One Raffles
Quay #18-00 South Tower Singapore 048583, +65 6423 8001), which may be contacted in respect of any matters
arising from, or in connection with, this report. Where this report is issued or promulgated in Singapore to a person who
is not an accredited investor, expert investor or institutional investor (as defined in the applicable Singapore laws and
regulations), they accept legal responsibility to such person for its contents.
Taiwan: Information on securities/investments that trade in Taiwan is for your reference only. Readers should
independently evaluate investment risks and are solely responsible for their investment decisions. Deutsche Bank
research may not be distributed to the Taiwan public media or quoted or used by the Taiwan public media without
written consent. Information on securities/instruments that do not trade in Taiwan is for informational purposes only and
is not to be construed as a recommendation to trade in such securities/instruments. Deutsche Securities Asia Limited,
Taipei Branch may not execute transactions for clients in these securities/instruments.
Qatar: Deutsche Bank AG in the Qatar Financial Centre (registered no. 00032) is regulated by the Qatar Financial Centre
Regulatory Authority. Deutsche Bank AG - QFC Branch may only undertake the financial services activities that fall
within the scope of its existing QFCRA license. Principal place of business in the QFC: Qatar Financial Centre, Tower,
West Bay, Level 5, PO Box 14928, Doha, Qatar. This information has been distributed by Deutsche Bank AG. Related
financial products or services are only available to Business Customers, as defined by the Qatar Financial Centre
Regulatory Authority.
Russia: This information, interpretation and opinions submitted herein are not in the context of, and do not constitute,
any appraisal or evaluation activity requiring a license in the Russian Federation.
Kingdom of Saudi Arabia: Deutsche Securities Saudi Arabia LLC Company, (registered no. 07073-37) is regulated by the
Capital Market Authority. Deutsche Securities Saudi Arabia may only undertake the financial services activities that fall
within the scope of its existing CMA license. Principal place of business in Saudi Arabia: King Fahad Road, Al Olaya
District, P.O. Box 301809, Faisaliah Tower - 17th Floor, 11372 Riyadh, Saudi Arabia.
United Arab Emirates: Deutsche Bank AG in the Dubai International Financial Centre (registered no. 00045) is regulated
by the Dubai Financial Services Authority. Deutsche Bank AG - DIFC Branch may only undertake the financial services
activities that fall within the scope of its existing DFSA license. Principal place of business in the DIFC: Dubai
International Financial Centre, The Gate Village, Building 5, PO Box 504902, Dubai, U.A.E. This information has been
distributed by Deutsche Bank AG. Related financial products or services are only available to Professional Clients, as
defined by the Dubai Financial Services Authority.
Australia: Retail clients should obtain a copy of a Product Disclosure Statement (PDS) relating to any financial product
referred to in this report and consider the PDS before making any decision about whether to acquire the product. Please
refer to Australian specific research disclosures and related information at
https://australia.db.com/australia/content/research-information.html
Australia and New Zealand: This research, and any access to it, is intended only for "wholesale clients" within the
meaning of the Australian Corporations Act and New Zealand Financial Advisors Act respectively.
Additional information relative to securities, other financial products or issuers discussed in this report is available upon
request. This report may not be reproduced, distributed or published without Deutsche Bank's prior written consent.
Copyright © 2016 Deutsche Bank AG

David Folkerts-Landau
Chief Economist and Global Head of Research
Raj Hindocha Marcel Cassard Steve Pollard

Global Chief Operating Officer Global Head Global Head
Research FICC Research & Global Macro Economics Equity Research
Michael Spencer Ralf Hoffmann Andreas Neubauer

Regional Head Regional Head Regional Head
Asia Pacific Research Deutsche Bank Research, Germany Equity Research, Germany
International locations
Deutsche Bank AG Deutsche Bank AG Deutsche Bank AG Deutsche Securities Inc.
Deutsche Bank Place Große Gallusstraße 10-14 Filiale Hongkong 2-11-1 Nagatacho
Level 16 60272 Frankfurt am Main International Commerce Centre, Sanno Park Tower
Corner of Hunter & Phillip Streets Germany 1 Austin Road West,Kowloon, Chiyoda-ku, Tokyo 100-6171
Sydney, NSW 2000 Tel: (49) 69 910 00 Hong Kong Japan
Australia Tel: (852) 2203 8888 Tel: (81) 3 5156 6770
Tel: (61) 2 8258 1234
Deutsche Bank AG London Deutsche Bank Securities Inc.
1 Great Winchester Street 60 Wall Street
London EC2N 2EQ New York, NY 10005
United Kingdom United States of America
Tel: (44) 20 7545 8000 Tel: (1) 212 250 2500
GRCM2016PROD035549

Telco Cloud Barclays Report

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Telco Cloud Barclays Report

Uploaded by

Copyright:

Available Formats

Deutsche Bank

Vijay Bhagavath, Ph.D

F.I.T.T. for investors (+1) 212 250-4711

Resistance is futile Ross Seymore

Disruption ahead for telecom equipment vendors

Europe Industry Date

Resistance is futile Research Analyst

Disruption ahead for telecom equipment vendors Johannes Schaller

Executive summary ............................................................. 3

Telcos: the story so far ........................................................ 8

Telco cloud: new paradigm ............................................... 14

Focus on mobile networks ................................................ 22

Implications for vendors .................................................... 30

Implications for coverage .................................................. 42

Implications for Telcos ...................................................... 48

Implications for ARM ........................................................ 51

Page 2 Deutsche Bank AG/London

The rising importance of the telco (network) cloud

Disruption ahead for equipment vendors

Deutsche Bank AG/London Page 3

Figure 3: Tech companies’ exposure to telcos (hardware, software, services)

Telco-exposure on total sales

Source: Deutsche Bank

C-RAN could be more impactful in mobile networks near-term

Page 4 Deutsche Bank AG/London

Figure 4: Likely impact of telco cloud on the wireless infrastructure market

2014-15 2016-17 2018-19 2020-23 2024-25

Proof of Early Early Broad Laggards

Key domains Mobile Core (vEPC, vIMS etc), C-RAN vRAN

Implications for coverage in network equipment

European coverage: we continue to prefer Nokia over Ericsson

Deutsche Bank AG/London Page 5

By contrast, Ericsson (Hold, SKr69 target), appears to be pushing a services-led

US networking equipment: we see a positive read for Cisco and Juniper

Bigger picture, we see Cisco and Juniper, in particular, increasingly leading

Vijay has written on the transition at networking companies to software in a

Telcos are set to benefit from network supplier disruption

Page 6 Deutsche Bank AG/London

The ARM processor architecture should benefit

Figure 5: Market share shifts in networking (by value)

2014 2016E 2018E 2020E

Intel is also set to see traction in networking

Deutsche Bank AG/London Page 7

Telcos: the story so far

Telcos: traffic growth is pressurizing profitability

Telcos are seeing fast-rising growth in IP traffic, driven primarily by mobile.

Figure 6: IP traffic by access technology, worldwide, exabytes per month

180 22% CAGR 2014-2019

Source: Cisco VNI Mobile, 2016

The increase in data growth necessitates huge capital investment by telcos,

Page 8 Deutsche Bank AG/London

Mobile operators - questionable payoff on 4G investments

Source: Deutsche Bank Source: Deutsche Bank

According to the Ericsson Mobility Report (Nov’15), Youtube accounts for

Next-generation technologies are essential to reduce cost

Deutsche Bank AG/London Page 9

The cloud has highlighted telcos’ lack of agility

Page 10 Deutsche Bank AG/London

Figure 11: Telco cycle Figure 12: Cloud cycle

Source: Telefonica Source: Telefonica

Cloud has already disrupted the IT environment

Deutsche Bank AG/London Page 11

2013 2014 2015 2016E 2017E

Source: Company financials, Deutsche Bank estimates